16134540

TESIS DOCTORAL
SINESTESIA Y EMOCIONES.
REACCIONES AFECTIVAS ANTE LA PERCEPCIN
DE ESTMULOS SINESTSICAMENTE INCONGRUENTES

Synesthesia and emotions.
Affective reactions as a consequence
of synesthetically incongruent stimuli

Doctoranda: Alicia Callejas Sevilla

Director: Juan Lupiez Castillo

Departamento de Psicologa Experimental y Fisiologa del Comportamiento

Universidad de Granada
Mayo 2006
Editor: Editorial de la Universidad de Granada
Autor: Alicia Callejas Sevilla
D.L.: Gr. 1171- 2006
ISBN: 84-338-3874-1

NDICES

ndice de Contenidos ________________________________________________________
5
NDICE DE CONTENIDOS
NDICES 3
NDICE DE CONTENIDOS _________________________________________________________ 5
NDICE DE FIGURAS, TABLAS Y GRFICAS ________________________________________ 9
SINESTESIA. INTRODUCCIN 15
1 Antecedentes histricos ___________________________________________________________ 18
2 Caractersticas de la sinestesia _____________________________________________________ 21
2.1 Estabilidad 21
2.2 Carcter perceptual 25
2.3 Carcter idiosincrsico 26
2.4 Automaticidad 28
2.5 Percepcin genrica 30
2.6 Carcter memorable de las percepciones 30
2.7 Unidireccionalidad 32
2.8 Carcter emocional 33
3 Prevalencia y carcter hereditario __________________________________________________ 34
4 Tipos de sinestesia _______________________________________________________________ 37
5 Bases cerebrales _________________________________________________________________ 39
6 Teoras Explicativas ______________________________________________________________ 41
6.1 Teoras Psicolgicas 41
6.2 Teoras Neurocognitivas 42
7 Conclusiones ____________________________________________________________________ 49
MOTIVATION OF THIS RESEARCH 51
OVERVIEW OF THE RESEARCH__________________________________________________ 54
SECTION 1: SYNESTHESIA AND ATTENTION 57
CHAPTER 1 AUTOMATIC PERCEPTION AND SYNAESTHESIA:
EVIDENCE FROM COLOUR AND PHOTISM NAMING IN A STROOP-
NEGATIVE PRIMING TASK 59
ABSTRACT______________________________________________________________________ 61
INTRODUCTION_________________________________________________________________ 61
EXPERIMENT 1. CONTROL EXPERIMENT_________________________________________ 64
METHOD 65
RESULTS 67
DISCUSSION 68
EXPERIMENTS 2a AND 2b. SYNAESTHESIA EXPERIMENTS_________________________ 68
METHOD 69
__________________________________________________________________ NDICE
6
RESULTS 71
DISCUSSION 74
CHAPTER 2 THE ROLE OF ATTENTION IN SYNESTHESIA 79
ABSTRACT______________________________________________________________________ 81
INTRODUCTION_________________________________________________________________ 81
EXPERIMENT 1. VISUAL SEARCH WITHOUT SEARCHING CUES____________________ 86
METHOD 87
RESULTS AND DISCUSSION 89
EXPERIMENT 2. VISUAL SEARCH WITH SEARCHING CUES ________________________ 91
METHOD 92
EXPERIMENTS 3a and 3b. VISUAL SEARCH WITH SEARCHING CUES AND
LIMITED EXPOSURE TIME_______________________________________________________ 95
METHOD 96
EXPERIMENT 4. VISUAL SEARCH WITH COLOR CUES____________________________ 100
METHOD 100
EXPERIMENT 5. ATTENTION MODULATION OVER PHOTISMS ELICITED BY
COMPLEX STIMULI ____________________________________________________________ 103
METHOD 105
GENERAL DISCUSSION _________________________________________________________ 114
SECTION 2: SYNESTHESIA AND EMOTION 119
CHAPTER 3 GREEN LOVE IS UGLY: EMOTIONS ELICITED BY
SYNESTHETIC GRAPHEME-COLOR PERCEPTIONS 121
ABSTRACT_____________________________________________________________________ 123
INTRODUCTION________________________________________________________________ 123
EXPERIMENTS 1a AND 1b. VALENCE JUDGMENT TASK___________________________ 126
METHOD 126
EXPERIMENTS 2a AND 2b. CATEGORIZATION TASK______________________________ 130
METHOD 131
CHAPTER 4 ARE ALL SYNESTHETES CREATED EQUAL ?
GENERALIZATIONS OF SYNESTHETICALLY INDUCED AFFECTIVE
REACTIONS 141
ABSTRACT_____________________________________________________________________ 143
ndice de Contenidos ________________________________________________________
7
INTRODUCTION________________________________________________________________ 143
EXPERIMENT 1a AND 1b. EFFECT OF PHOTISM-COLOR CONGRUENCY IN A
CATEGORIZATION TASK._______________________________________________________ 145
METHOD 145
EXPERIMENTS 2a AND 2b. TEST OF THE STRENGTH OF PSVs
SYNESTHETIC EXPERIENCES: STROOP INTERFERENCE WITH A COLOR
NAMING AND A PHOTISM NAMING TASK________________________________________ 155
METHOD 156
EXPERIMENT 3. CATEGORIZATION OF COLOR ACCURACY ______________________ 162
METHOD 163
EXPERIMENT 4. CONTROL EXPERIMENT FOR EXPERIMENT 3____________________ 166
METHOD 166
CHAPTER 5 CONTEXT EFFECTS ON SYNESTHETICALLY INDUCED
AFFECTIVE REACTIONS 173
ABSTRACT_____________________________________________________________________ 175
INTRODUCTION________________________________________________________________ 175
EXPERIMENT 1. EFFECT OF COLORING WORDS IN BLACK IN A COLOR-
CONTEXT TASK________________________________________________________________ 177
METHOD 178
EXPERIMENT 2a AND 2b. EFFECT OF COLORING WORDS IN GREY IN A
COLOR-CONTEXT TASK AND AN ACHROMATIC-CONTEXT TASK_________________ 182
METHOD 182
EXPERIMENT 3. SUBJECTIVE EXPERIENCE OF COLOR VS. COLOR WAVE
LENGHT _______________________________________________________________________ 189
METHOD 189
CHAPTER 6 SYNESTHETICALLY INDUCED AFFECTIVE REACTIONS
AND ATTITUDE FORMATION 197
ABSTRACT_____________________________________________________________________ 199
INTRODUCTION________________________________________________________________ 199
EXPERIMENT 1. IMPLICIT ATTITUDE FORMATION WITH IMPLICIT
SYNESTHETIC INCONGRUENCE AS THE UNCONDITIONED STIMULUS.____________ 200
METHOD 201
__________________________________________________________________ NDICE
8
EXPERIMENT 2a. IMPLICIT ATTITUDE FORMATION WITH EXPLICIT
SYNESTHETIC INCONGRUENCE AS UNCONDITIONED STIMULUS_________________ 208
METHOD 208
EXPERIMENT 2b. SECOND SESSION _____________________________________________ 212
METHOD 212
CHAPTER 7 ARE SYNESTHETES THE END OF A CONTINUOM?
COLOR EMOTION RELATIONS IN NON-SYNESTHETES 217
ABSTRACT_____________________________________________________________________ 219
INTRODUCTION________________________________________________________________ 219
EXPERIMENT 1. VALENCE CATEGORIZATION AS A FUNCTION OF COLOR
PREFERENCE AND SUBJECTIVE COLOR LIGHTNESS_____________________________ 221
METHOD 222
EXPERIMENT 2. FOLLOW UP____________________________________________________ 233
METHOD 233
RESUMEN DE RESULTADOS 245
GENERAL DISCUSSION 257
SUMMARY OF FINDINGS________________________________________________________ 259
SYNESTHESIA AND ATTENTION_________________________________________________ 266
1 How automatic is synesthesia? 266
2 Can synesthesia be suppressed? 268
3 Is synesthesia preattentional? 270
SYNESTHESIA AND EMOTION___________________________________________________ 271
1 Are there affective reactions associated to the experience of synesthesia? 272
2 Are these affective reactions automatic? Can they be ignored? 273
3 Are affective reactions a ubiquitous phenomenon? 275
4 Can these affective reactions be modified or extinguished? 276
5 Are synesthetic affective reactions strong enough as to condition co-occurring events? 278
6 Proposal of the bidimensional assessment hypothesis 279
FUTURE DIRECTIONS __________________________________________________________ 282
Individual differences 282
Factors influencing synesthetic affective reactions 283
Physiological measures of synesthetically elicited affective reactions 283
Similarities between synesthetes and non-synesthetes 284
CONCLUSSIONS ________________________________________________________________ 284
REFERENCES 287

ndice de Figuras, Tablas y Grficas_____________________________________________
9
NDICE DE FIGURAS, TABLAS Y GRFICAS
Grfica 1. Evolucin del nmero de publicaciones sobre sinestesia en los ltimos 150
aos. Datos procedentes de Marks (1975), Harrison (2001) y de las bases de datos de
Science Citation Index y Social Sciences Citation Index (bsqueda synesthesia or
synaesthesia) ____________________________________________________________________ 20
Figura 1. Ejemplo de consistencia en el color elegido por MA para cada letra y
nmero en cinco ocasiones distintas: las tres primeras columnas se rellenaron con la
paleta de color (40 colores) de Microsoft Excel. 1: inmediatamente despus de decir la
etiqueta verbal, 2: a los cinco minutos y 3: a la semana. Las dos ltimas columnas
corresponden a un retest 3 aos ms tarde. Los colores se eligieron de una muestra de
ms de 92000 colores en dos momentos consecutivos (4: primera pasada, 5: segunda
pasada).__________________________________________________________________________ 23
Figura 2. Ejemplo de colores para tres personas con sinestesia grafema-color. Como
puede observarse, uno de los casos solo percibe color para las vocales. ______________________ 26
Figura 3. Comparacin de los colores utilizados en un mtodo de aprendizaje de
lectura (Cartilla Micho), presentados a la izquierda, con los colores asociados a las
letras por la participante sinestsica MA, presentados a la derecha. ________________________ 28
Figura 4. Representacin esquemtica de la teora de la activacin cruzada de
Ramachandran y Hubbard, 2001- ____________________________________________________ 44
Figura 5. Representacin esquemtica de la teora de la retroalimentacin desinhibida
de Grossenbacher y Lovelace (2001).__________________________________________________ 46
Figura 6. Representacin esquemtica de la teora del procesamiento reentrante de
Smilek, Dixon y Merikle (2001) y comparacin con la teora de activacin cruzada de
Ramachandran y Hubbard (2001b). __________________________________________________ 48
Table I. Mean RTs (in ms) and error percentages (in parenthesis) for each
experimental condition of the Control Experiment.______________________________________ 67
Table II. Mean RTs (in ms) and error percentages (in parenthesis) for each
experimental condition for the Control Group Colour-Naming Experiment, and for
MA Colour-Naming and Photism-Naming Experiments. _________________________________ 71
Graph 1: Stroop effects in MA's performance, in the colour-naming and photism-
naming experiments, as compared to the control group of non-synaesthetic
participants. For comparison, data from the Control Experiment are also provided. __________ 72
Graph 2. Negative Priming effect in MA's performance, in the colour-naming and
photism-naming experiments, as compared to the control group of non-synaesthetic
participants. For comparison, data from the Control Experiment are also provided. __________ 73
Figure 1. Experiment 1. Three examples of target display. In the left one the target is a
vertical rectangle made out of Ns; in the middle one the horizontal rectangle is
made out of Ws and in the right one the horizontal rectangle is made out of Ms.__________ 88
Table I. Experiment 1. RT (and percentage correct) per group and experimental
condition. ________________________________________________________________________ 89
Graph 1. Experiment 1. Reaction times for MA and the Control group as a function of
the letter that formed the target stimulus. Error bars denote standard errors of the
mean.____________________________________________________________________________ 90
__________________________________________________________________ NDICE
10
Graph 2. Experiment 1. Percentages of correct responses for MA and the Control
group as a function of the letter that formed the target stimulus. Error bars denote
standard errors of the mean. ________________________________________________________ 91
Table II. Experiment 2. RT (and error rates) per group and experimental condition.__________ 92
mean.____________________________________________________________________________ 93
Graph 4. Experiment 2. Error percentages for MA and the Control group as a
function of the letter that formed the target stimulus. Error bars denote standard
errors of the mean. ________________________________________________________________ 94
Table III. Experiments 3a and 3b. Error rates per group and experimental condition._________ 96
Graph 5. Experiment 3a and 3b. Error rates for MA and the Control group as a
function of the time the display was presented and letter that formed the target
stimulus. Error bars denote standard errors of the mean. ________________________________ 97
horizontal rectangle made out of Ws; in the middle one the vertical rectangle is
made out of Ns and in the right one the horizontal rectangle is made out of Ms. _________ 100
Table IV. Experiment 4. RT (and percentage of errors) per group and experimental
condition. _______________________________________________________________________ 101
mean.___________________________________________________________________________ 102
Graph 7. Experiment 4. Error rates for MA and the Control group as a function of the
letter that formed the target stimulus. Error bars denote standard errors of the mean. _______ 103
Table V. Experiment 5. Experimental design and items used with the corresponding
elicited photism. Only the Attend Global condition is shown since the stimuli were the
same and only instructions to attend to the global or local form varied between
conditions. ______________________________________________________________________ 106
Table VI. Experiment 5. RT per group and experimental condition._______________________ 108
Graph 8. Experiment 5. Reaction times for MA and the Control group when
responding to pure-level conditions. Error bars denote standard errors of the mean _________ 109
Graph 9a and 9b. Experiment 5. Reaction times for MA and the Control group when
responding to inconsistent-letter conditions. a) Conditions where the photism of the
local letter was incongruent with the displayed color (collapsed for global letter
congruency). b) Conditions where the photism of the global letter was incongruent
with the displayed color (collapsed for local letter congruency). Error bars represent
standard errors of the mean. _______________________________________________________ 112
Graph 10. Experiment 5. Reaction times for MA and the Control group as a function
of the congruency of the attended level. Error bars denote standard errors of the
mean.___________________________________________________________________________ 113
Graph 11. Experiment 5. Reaction times for MA and the Control group as a function
of Global and Local Photism-Color Congruency. Error bars denote standard errors of
the mean. _______________________________________________________________________ 114
11
Graph 1. Experiment 1a and 1b. Congruency effect as a function of group. The black
bar shows the results of Experiment 1b where all words were presented in black.
Error bars denote standard error of the mean. ________________________________________ 128
Graph 2. Experiment 1a and 1b. Valence ratings. Mean valence ratings for control
group, and M.A. Error bars represent the standard error of the mean. ____________________ 129
Graph 3. Experiment 2a and 2b. A). Mean reaction time data. B). Mean error rate
data. Error bars represent the standard error of the mean. ______________________________ 133
Table I. Experiments 1a and 1b. Mean categorization RT per group, subject and
condition. Error percentages are shown in parenthesis. _________________________________ 148
Graph 1. Experiment 1a. Reaction times for each Group and Case as a function of
Congruence. _____________________________________________________________________ 149
Graph 2. Experiment 1a and 1b. Reaction times for each Group as a function of
Valence and Congruence. __________________________________________________________ 150
Graph 3. Experiment 1a and 1b. Reaction times for each synesthete as a function of
Graph 4. Experiment 1a. Error percentage for each Group as a function of Valence
and Congruence. _________________________________________________________________ 152
Graph 5. Experiment 1a. Percentage correct for each synesthete case as a function of
Table II. Experiment 2a (color naming) and 2b (photism naming). Mean naming times
per condition for each group and experiment. Error rates are shown in parenthesis. _________ 158
Graph 6. Experiment 2a and 2b. Stroop effect. Reaction times for PSV for the Color
Naming and Photism Naming version of the Stroop task and reaction times for her
Control group for the Color Naming version.__________________________________________ 158
Graph 7. Experiment 2a and 2b. Negative Priming effect. Reaction times for PSV for
the Color Naming and Photism Naming tasks and Color Naming for her Control. ___________ 159
Table III. Experiment 3. Mean categorization times per synesthete and condition.
Error rates are shown in parenthesis ________________________________________________ 164
Graph 8. Experiment 3. Reaction times for PSV and MA as a function of word
Valence and Color Congruence._____________________________________________________ 165
Table IV. Experiment 4. Mean categorization times per condition. Error rates shown
in parenthesis ____________________________________________________________________ 167
Graph 9. Experiment 4. Reaction times for the Control group as a function of word
Valence and Color Congruence._____________________________________________________ 168
Graph 1. Results from Experiment 1a and 1b of Callejas et al. (submitted, Chapter 3).
Table I. Experiment 1. Mean valence judgment per group and condition. __________________ 179
Graph 2. Experiment 1. Valence rating for congruent, incongruent and black colored
words as a function of group and valence. Error bars denote standard error of the
mean.___________________________________________________________________________ 180
__________________________________________________________________ NDICE
12
Graph 3. Experiment 1 compared to Experiment 1b from Callejas et al. (submitted,
Chapter 3). Ratings for black colored words as a function of color Context and
Valence. ________________________________________________________________________ 181
Figure 1. Experiment 2a and 2b. Stimuli and procedure used. ____________________________ 183
Table II. Experiment 2a. Mean valence judgment per group and condition. ________________ 184
Graph 4. Experiment 2a. Valence rating for congruent, incongruent and grey colored
words as a function of group and valence. Error bars denote standard error of the
mean.___________________________________________________________________________ 184
Table III. Experiment 2b. Mean valence judgment per group and condition.________________ 185
Graph 5. Experiment 2b. Mean valence judgments for MA and the Control group
when words were all presented in grey color. Error bars denote standard error of the
mean.___________________________________________________________________________ 186
Graph 6. Experiment 2a and Experiment 2b for MA. Ratings for grey colored words
as a function of color Context and Valence. ___________________________________________ 187
Graph 7. Comparison of results from Experiment 1b (Callejas et al., submitted,
Chapter 3), Experiment 1, 2a and 2b. Ratings for Black and Grey words as a function
of Valence and color Context._______________________________________________________ 187
Figure 2. Experiment 3. Stimuli and procedure used. ___________________________________ 190
Table IV. Experiment 3. Mean valence judgment per group and condition. _________________ 190
Graph 8. Experiment 3. Mean valence judgments for MA and the Control group when
words were all presented in subjectively perceived white color. Error bars denote
standard error of the mean. ________________________________________________________ 191
Graph 9. Experiment 2b and 3. Mean valence judgments for MA and the Control
group as a function of Valence and subjectively perceived color. Error bars denote
standard error of the mean. ________________________________________________________ 191
Table I. MAs colors for numbers, conditioned images and sample of distracters.____________ 203
Table II. Experiment 1. Mean RT and error rate for Controls and MA as a function of
Congruence and Distracter number. _________________________________________________ 205
Table III. Experiment 1. Valence ratings for CS- and CS+ and percentage of positions
remembered for each control and for MA. ____________________________________________ 206
Table IV. Experiment 2. Mean RT and error rate for Controls and MA as a function
of Congruence and Distracter number. _______________________________________________ 210
Table V. Experiment 2. Valence ratings for CS- and CS+ and percentage of positions
remembered for each control and for MA. ____________________________________________ 211
Table VI. Experiment 2b. Valence ratings for CS- and CS+ before and after the
conditioning phase and percentage of positions remembered after the conditioning
phase for MA. ___________________________________________________________________ 214
Table I. Experiment 1. Mean reaction times and error rates (in parenthesis) per
experimental condition for the general analysis. _______________________________________ 224
Graph 1. Experiment 1. Reaction times as a function of valence and location. Error
bars denote standard error of the mean. ______________________________________________ 225
13
Graph 2. Experiment 1. Reaction times as a function of color, valence and location.
Graph 3. Experiment 1. Mean number of votes per color for Subjective Lightness and
Subjective Preference. _____________________________________________________________ 227
Table II. Experiment 1. Mean reaction times and error rates (in parenthesis) per
experimental condition for the Lightness analysis and the Preference Analysis. _____________ 227
Graph 4. Experiment 1. Lightness analysis. Mean RTs for light and dark colors as a
function of position and valence. Error bars denote standard error of the mean. ____________ 228
Graph 5. Experiment 1. Preference analysis. Mean RTs for liked and disliked colors as
a function of position and valence. ___________________________________________________ 229
Graph 6. Experiment 1. Preference analysis. Mean error percentage for liked and
disliked colors as a function of position and valence. Error bars denote standard error
of the mean. _____________________________________________________________________ 231
Table III. Experiment 2. Mean reaction times and error rates (in parenthesis) per
experimental condition for the general analysis. _______________________________________ 234
Graph 7. Experiment 2. Reaction times as a function of valence and location. Error
bars denote standard error of the mean. ______________________________________________ 234
Graph 8. Experiment 2. Reaction times as a function of color, valence and location.
Subjective Preference. Error bars denote standard error of the mean. _____________________ 236
Table IV. Experiment 2. Mean reaction times and error rates (in parenthesis) per
experimental condition for the Lightness analysis and the Preference Analysis. _____________ 236
function of position and valence. Error bars denote standard error of the mean. ____________ 237
Graph 11. Experiment 2. Preference analysis. Mean RTs for liked and disliked colors
as a function of position and valence. Error bars denote standard error of the mean._________ 238
disliked colors as a function of position and valence. Error bars denote standard error
of the mean. _____________________________________________________________________ 240
Figure 1. Depiction of the Bidimensional Assessment Hypothesis proposed to explain
the affective reactions found in synesthetes when perceiving a stimulus colored
congruently or incongruently with their internal experience. _____________________________ 281

SINESTESIA. INTRODUCCIN
1

1
Partes de este captulo han sido publicadas en:
Callejas, A. y Lupiez, J. (en prensa). Los colores de mis letras: sinestesia grafema-color. En M.J. Contreras, J.
Botella, R. Cabestrero y B. Gil (Coords.), Lecturas de Psicologa Experimental. Madrid: UNED.
Callejas, A. y Lupiez, J. (2005). El color de las palabras. Percepnet. Centro de recursos sobre la percepcin y
ciencias sensoriales. Revista electrnica. http://www.percepnet.com/
Introduccin ______________________________________________________________
17
Normalmente vemos los colores, olemos los olores, y omos los sonidos. Pero,
por raro que parezca, esto no es siempre as, como ocurre, al parecer, en el caso de
los bebs recin nacidos (Maurer, 1997) o en las personas ya adultas que
experimentan el fenmeno de la percepcin sinestsica. La sinestesia, del griego syn
(unin) +aisthesis (sensacin), consiste en la unin de distintos sentidos. As, a la
percepcin de un estmulo sensorial se une la experiencia subjetiva de otra percepcin
sin referente externo (Cytowic, 1989). Una experiencia es evocada por un estmulo
especfico que no evocara dicha experiencia en la mayora de la poblacin (Ward y
Mattingley, 2006). En algunos casos la percepcin de una meloda va acompaada de
una experimentacin de distintos sabores (Beeli, Esslen y J ncke, 2005), la
percepcin de distintos sabores va acompaada de una experiencia de tacto (Cytowic,
1993), o cada palabra se experimenta con un sabor particular (Ward y Simner, 2003).
En todos estos casos hay un estmulo, perceptual o conceptual, que evoca o dispara
una sensacin adicional y esa sensacin que se experimenta aadida a la propia del
estmulo. En la literatura cientfica se denomina al estmulo evocador inductor y a la
sensacin adicional experimentada concurrente (Grossenbacher y Lovelace, 2001).
Aunque el trmino original de sinestesia hace referencia a la unin de
diferentes sentidos, existen casos en los que sta se produce dentro de una misma
modalidad sensorial pero entre dimensiones estimulares distintas como el color y la
forma. As, aunque existen muchos tipos distintos de sinestesia, el ms comn, y tal
vez por ello el ms estudiado, es aquel en el que las letras, palabras o nmeros
evocan colores (Day, 2005). Este tipo de sinestesia se ha denominado grafema-color
haciendo referencia al inductor (el grafema) y al concurrente (el color). En segundo
lugar estn las personas que perciben colores cuando se enfrentan a unidades de
tiempo. Tambin es frecuente el caso de ver colores para las palabras habladas, los
sonidos en general o las notas musicales. En menor medida se dan casos de
personas que ven colores para distintos sabores, sabores o percepciones tctiles para
distintos sonidos, sabores para percepciones tctiles y un largo etctera. Hasta el
momento se han documentado casos de casi todas las distintas combinaciones
posibles entre los distintos sentidos (Day, 2005; Rich, Bradshaw y Mattingley, 2005).
Es importante distinguir la sinestesia como fenmeno perceptual de otros
fenmenos como las asociaciones basadas en memoria o de la sinestesia como figura
literaria. En el caso de las asociaciones, un estmulo determinado puede estar
asociado con un color particular. Cuando se le pregunta a una persona lo que le
sugiere la palabra pltano es normal que responda que le sugiere una fruta o el color
______________________________________________ SINESTESIA. INTRODUCCIN
18
amarillo. No obstante, no perciben la palabra como si fuera amarilla sino que la ven
escrita del color en que realmente est escrita. De igual modo, la sinestesia como
figura literaria se utiliza para embellecer el estilo de una composicin, para enriquecer
la descripcin de un fenmeno (ej. Mara es una persona muy dulce llevaba una
camiseta naranja chilln) pero ello es muy distinto del hecho de que el escritor est
experimentando tales percepciones.
Al contrario, un sinestsico que experimenta colores para las palabras, puede
que vea la palabra pltano de color amarillo, o tal vez la vea de color azul, pero la ve
realmente de ese color aunque est escrita con tinta negra; tiene una experiencia
subjetiva similar a la experimentada ante la percepcin real de colores. O puede que
un sinestsico realmente experimente un sabor dulce al ver a una persona particular.
Los sinestsicos grafema-color perciben las letras y nmeros como si estuvieran
escritas con un color determinado. No obstante, son conscientes de que ese no es el
color real del estmulo, o al menos, de que no es la nica experiencia de color
proveniente de ese estmulo sino que experimentan dos colores, real y sinestsico. Es
interesante destacar el hecho de que no todos los sinestsicos de esta categora
perciben los colores de igual modo. Algunos de ellos ven el color de las letras
proyectado hacia el exterior de manera que se superpone a la letra en s o flota en el
campo visual sin localizarse especficamente en el espacio que ocupa la letra,
mientras que otros lo ven en la mente (Dixon, Smilek y Merikle, 2004).
1 Antecedentes histricos
El fenmeno de la sinestesia ha sido conocido durante mucho tiempo aunque
su estudio cientfico es relativamente reciente. Ya en el siglo XVIII se encuentran
escritos como el de Castel (Castel, 1735) en el que se hace referencia a estudios
previos sobre un caso de sinestesia en una persona ciega. Por estas mismas fechas
Isaac Newton (1730) intent encontrar la frmula matemtica para igualar la vibracin
de las ondas sonoras a la longitud de onda correspondiente para diferentes colores.
Con posterioridad, ms de un centenar de escritos fueron publicados sobre sinestesia
en los que se haca especial hincapi en la escucha coloreada o en las vocales
coloreadas (vase Marks, 1975 para una revisin de este material).
En 1895 Mary W. Calkins public el primer artculo cientfico en el que la
palabra sinestesia apareca en el ttulo y se utilizaba con el significado que le
atribuimos actualmente. Un par de aos antes realizara un novedoso estudio con el fin
Introduccin _____________________________________________________________
19
de estimar la frecuencia de dicho fenmeno. Como profesora de psicologa del Colegio
Wellesley tom un grupo de 525 alumnas a las que cuestion en relacin a sus
percepciones. Un grupo de 35 result experimentar lo que en aquel momento Calkins
denomin pseudocromatestesia y que ms tarde llamara sinestesia. Por lo tanto el
6.7% pareca experimentar alguna forma de sinestesia. Esta cifra es bastante superior
a cualquiera de las estimaciones propuestas actualmente. No obstante, tras examinar
las respuestas de las alumnas encontr que la proporcin de estudiantes que
experimentaba color para las palabras no era superior al 1%. Esta cifra s es ms
cercana a otras estimaciones para la sinestesia grafema-color ms recientes (vase
seccin de prevalencia).
Durante esta poca tambin se publicaron diversos libros en los que se discuta
la sinestesia en el mbito del arte, la msica, la literatura, la lingstica o la filosofa
(Surez de Mendoza, 1890; Argelander, 1927). En la mayora de los casos se
centraban en la escucha coloreada como la forma ms frecuente de sinestesia. En
este primer cuarto del siglo XX segua vivo el inters por la sinestesia y se dieron a
conocer numerosos casos de escucha coloreada, as como de msica coloreada o de
vocales con color. Incluso se intent explicar la sinestesia desde el punto de vista de la
teora psicoanaltica (Hug-Hellmuth, 1912) o segn la estructura fontica de las
vocales (Henning, 1923).
En los cincuenta aos posteriores, probablemente por la influencia del
conductismo y la negacin de toda conducta no observable, el inters por la sinestesia
pareci decaer, aunque todava fueron numerosos los estudios realizados, que
incluan algunos sobre msica coloreada (Karwoski, Odbert y Osgood, 1942; Odbert,
Karwoski y Eckerson, 1942), intentos de estudiar las similitudes entre los colores
experimentados por las vocales en distintos idiomas (Reichart, J akobson y Perth,
1949) o informes sobre escucha coloreada inducida por sustancias como la mescalina
(Delay, Grard y Racamier, 1951). Como puede observarse en la Grfica 1, el nmero
de producciones en esta poca descendi considerablemente, comparado con los
cincuenta aos anteriores.
En la dcada de los ochenta, Cytowic y Wood (1982) realizaron una serie de
observaciones sistematizadas de una persona que experimentaba formas con los
sabores y otra que escuchaba colores. Este trabajo puede considerarse el primer
estudio experimental, al menos desde el renacimiento del inters por el fenmeno de
la sinestesia.
20
A partir de este momento, y gracias al surgimiento gran auge de la psicologa y
la neurociencia cognitiva, han sido numerosos los estudios interesados por el
fenmeno de la sinestesia. Principalmente se han centrado en demostrar que es un
fenmeno genuino y no una simple forma de expresarse (Baron-Cohen, Wyke y
Binnie, 1987; Mills, Boteler y Oliver, 1999; Dixon, Smilek, Cudahy y Merikle, 2000;
Ramachandran y Hubbard, 2001a; Palmeri y cols., 2002; Lupiez y Callejas, 2006)
aunque tambin ha habido un gran inters por la posibilidad de que la sinestesia sea
hereditaria (Baron-Cohen et al., 1996; Ward y Simner, 2005) y en los mecanismos
cerebrales implicados en este fenmeno (Aleman y cols, 2001; Hubbard, Arman,
Ramachandran y Boynton, 2005; Nunn y cols., 2002; Paulesu y cols., 1995; Weiss y
cols., 2001).
Evolucin del Nmero de publicaciones sobre sinestesia
1850-2006
0
20
40
60
80
100
120
140
1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
Ao
N
m
e
r
o

d
e

p
u
b
l
i
c
a
c
i
o
n
e
s

Grfica 1. Evolucin del nmero de publicaciones sobre sinestesia en los ltimos 150
aos. Datos procedentes de Marks (1975), Harrison (2001) y de las bases de datos de
Science Citation Index y Social Sciences Citation Index (bsqueda synesthesia or
synaesthesia )
Una vez asentada la premisa de que la sinestesia es un fenmeno real,
cuantificable y sujeto a estudio emprico, se ha comenzado a diversificar el inters de
la comunidad cientfica para centrarse en otros tipos de sinestesia menos conocidos,
aunque tambin relativamente frecuentes, como pueden ser las lneas numricas
(Seron y cols., 1992), la msica coloreada (Ward, Huckstep y Tsakanikos, 2006), la
sinestesia lxico-gustativa (Ward y Simner, 2003), la msica saboreada (Beeli y otros,
2005), la sinestesia crono-espacial (Smilek, Callejas, Dixon y Merikle, en prensa), etc.
Actualmente el fenmeno de la sinestesia parece estar de nuevo en auge y la
evolucin en los ltimos aos ha sido espectacular. No slo hay ms laboratorios
Introduccin _____________________________________________________________
21
interesados en desentraar las peculiaridades de esta forma de percibir, y de cmo
puede ayudarnos a conocer ms a fondo el funcionamiento y las bases neurales de los
procesos perceptuales y de integracin multisensorial en personas no sinestsicas,
sino que tambin ha crecido la conciencia social del fenmeno que se encontraba en
el nivel de lo paranormal o anecdtico. Numerosos libros se han publicado en los
ltimos veinte aos, tanto a nivel divulgativo (Cytowic, 1993; Harrison, 2001, Duffy,
2001) como a nivel cientfico (Cytowic, 1989/2002; Baron-Cohen y Harrison, 1997;
Robertson y Savig, 2005). Se ha creado una asociacin de sinestesia en Estados
Unidos (American Synesthesia Association) y otra paralela en el Reino Unido (UK
Synaesthesia Association) y se celebran congresos anuales en ambos paises en los
que la comunidad cientfica interesada en la sinestesia, as como personas con
sinestesia, se renen y ponen en comn los avances cientficos junto con los informes
subjetivos de los sinestsicos. En el mbito puramente cientfico tambin se ha
plasmado el inters creciente en el tema con un simposium especial sobre sinestesia
en el XIV congreso de la Sociedad Europea de Psicologa Cognitiva (European Society
of Cognitive Psychology) y una edicin especial de la revista Cortex dedicada
exclusivamente al tema de la sinestesia.
2 Caractersticas de la sinestesia
Aunque hay cierto debate sobre algunas de las afirmaciones siguientes, las
caractersticas ms importantes de la sinestesia como fenmeno neurocognitivo son
las siguientes

(Cytowic, 1995). La sinestesia es un fenmeno estable, de carcter
perceptual e idiosincrsico. Ocurre de manera automtica y es una experiencia
genrica memorable. Es unidireccional y tiene un marcado carcter emocional. A
continuacin se describen cada uno de estos aspectos.
2.1 Estabilidad
La sinestesia es un fenmeno estable en el tiempo. Esto es algo que lo
distingue de las asociaciones de memoria. Antes de que la sinestesia se empezara a
estudiar cientficamente, se solan minusvalorar las informaciones proporcionadas por
los sinestsicos sobre la base de que sus informes no estaban indicando fenmenos
reales sino meras asociaciones aprendidas a lo largo de la vida. As, algunos
argumentaban que las personas con sinestesia simplemente haban aprendido a leer y
escribir utilizando un mtodo en el que se asociaba un color a cada letra. Aunque esto
pueda ser el desencadenante de los colores particulares (Witthoft y Winawer, 2006)
22
que una persona experimenta, el hecho de que los experimente no est basado en
tales factores. Estudios realizados posteriormente han demostrado que sta no es la
razn de tales experiencias y que no se clasifican como asociaciones de conceptos
sino como percepciones. Cuando se pregunta a una persona con sinestesia sobre los
colores que ve para un conjunto de estmulos, la respuesta que se obtiene es mucho
ms estable que la que se pueda obtener preguntando a un grupo de personas sin
sinestesia. Esto es as incluso cuando el intervalo entre test y retest, en los no
sinestesicos, es de das y en los sinestsicos de meses (Baron-Cohen, Wyke y Binnie
(1987). La constancia en los colores percibidos por las personas sinestsicas es
sorprendente, ya que de 100 palabras, letras o nmeros que se les presenten,
responden seleccionando el mismo matiz de una paleta de colores para ms de 90 de
ellas. Es ms, aquellas en las que no responden con el mismo color, el escogido suele
ser uno del mismo grupo (ej. un azul ms claro o ms oscuro).
La estabilidad de las percepciones se ha convertido en un prerrequisito para la
sinestesia y, aunque algunos autores proponen entenderlo como una caracterstica
asociada ms que como una caracterstica definitoria (Ward y Mattingley, 2006),
cualquier estudio sobre sinestesia publicado en la actualidad presenta medidas de
estabilidad test re-test como forma de demostrar a priori que los participantes tienen de
hecho sinestesia. Tanto es as que ha sido desarrollado un Test de Autenticidad de la
sinestesia (Testo of Genuineness) (Baron-Cohen y cols, 1987; Asher y cols., 2006). Un
conjunto de fichas con distintos colores se presentan a los participantes para que elijan
el color que ms se parece a su percepcin para determinados estmulos (palabras o
sonidos). En una segunda ocasin se presentan los mismos estmulos y, con un
sistema de puntuacin determinado, se estima cmo de similares son los colores
elegidos para el mismo estmulo en dos ocasiones diferentes. Normalmente esta
prueba se realiza con un intervalo entre test-retest de meses o aos y sus autores
encuentran que las personas con sinestesia escogen el mismo color o uno muy
parecido en la mayora de los casos mientras que participantes no sinestsicos no
recuerdan ms del 30% de los colores elegidos incluso cuando el intervalo entre test-
retest no es mayor de unas cuantas semanas. En la Figura 1 puede observarse el
color elegido por una persona con sinestesia, evaluada en nuestro laboratorio, cuando
se le pidi que escogiera un color para cada letra y nmero de entre los 40 colores de
la paleta presente en Microsoft Excel. Como puede observarse, la consistencia entre
distintos momentos es muy alta y, slo en algunos casos, se encuentran diferencias en
el color elegido, aunque siempre dentro del mismo matiz. En estos casos la
participante inform que ninguna de las opciones de color ofrecidas era la adecuada y,
Introduccin _____________________________________________________________
23
por eso, present una mayor variabilidad en la eleccin del mismo. Tambin se
presentan los datos de un retest llevado a cabo tras tres aos con un sistema mucho
ms exhaustivo que daba opcin a elegir entre ms de 92.000 colores.
Letras Color 1 2 3 4 5
A blanco
B rojo
C marrn
D rojo
E rojo
F amarillo
G naranja
H blanca
I azul claro
J verde oliva
K marrn
L azul
LL azul
M amarillo
N amarillo
naranja
O amarillo
P naranja
Q marrn
R negro
S azul
T negro
U rosa
V violeta-rosa
W violeta-rosa
X violeta-azul
Y negra
Z marrn
Nmero Color 1 2 3 4 5
0 blanco
1 amarillo
2 rojo
3 verde
4 naranja
5 rojo
6 azul
7 amarillo
8 marron
9 azul marino
10 amarilo
11 amarillo
12 rojo
13 verde
14 naranja
15 marron
16 azul
17 amarillo
18 marron
19 azul marino
20 amarillo
Figura 1. Ejemplo de consistencia en el color elegido por MA para cada letra y nmero en
cinco ocasiones distintas: las tres primeras columnas se rellenaron con la paleta de
color (40 colores) de Microsoft Excel. 1: inmediatamente despus de decir la etiqueta
verbal, 2: a los cinco minutos y 3: a la semana. Las dos ltimas columnas corresponden
a un retest 3 aos ms tarde. Los colores se eligieron de una muestra de ms de 92000
colores en dos momentos consecutivos (4: primera pasada, 5: segunda pasada).
No obstante, estudios recientes apuntan a que la sinestesia no es un fenmeno
tan estable como se haba pensado. En un anlisis de un cuestionario, cumplimentado
24
por ms de 300 personas con sinestesia de habla inglesa, se encontr que un 30% de
los sinestsicos informaban de que sus experiencias sinestsicas se haban
modificado a lo largo del tiempo (Callejas, Smilek, Dixon & Merikle, 2004). De este
subgrupo un 32% inform de que el concurrente experimentado haba cambiado en
intensidad y otro 24% de que haba cambiado cualitativamente. Por lo tanto, la
ausencia de estabilidad test-retest, cuando el intervalo de tiempo es suficientemente
amplio, no debera de tomarse como prueba de ausencia de la condicin.
Para evitar este problema otros laboratorios han desarrollado mtodos con los
que se puede diferenciar claramente a las personas con sinestesia del resto de forma
muy rpida. Con una aplicacin informtica Smilek, Dixon y Merikle (2005b,
www.synaesthesia.uwaterloo.ca/participation.htm) presentan a los participantes una
matriz de color en la que el nmero de posibles colores supera los 92000 y, tras
asignar un color a cada letra y nmero, vuelven a presentar todos los estmulos en un
orden aleatorio distinto. Con este mtodo son capaces de distinguir, de forma clara, las
personas con sinestesia de aquellas que informan no percibir colores para letras o
nmeros.
En nuestro laboratorio hemos adaptado el sistema de Smilek y cols. (2005b) al
alfabeto espaol (inclusin de la letra ) y estamos recogiendo datos de multitud de
personas con sinestesia de habla hispana (www.ugr.es/~sinestes/). Los participantes
asignan un color de una amplia gama a cada letra o nmero presentados de forma
aleatoria. Inmediatamente despus de responder a los 37 estmulos, estos son
presentados de nuevo en otro orden aleatorio. Para evitar que los participantes puedan
obtener una consistencia mayor mediante la memorizacin de las coordenadas
espaciales en las que se encuentra cada color, una vez terminada la primera fase se
modifica el espacio de color presentado para que ningn color se encuentre en la
misma pareja de coordenadas x,y. Una vez terminada la segunda fase, los
participantes informan sobre el nivel de exactitud con que encontraron el color
experimentado para cada estmulo. En la actualidad estamos desarrollando un sistema
similar para evaluar la consistencia de los colores asociados a palabras.
Dando un paso ms, Eagleman (www.synesthete.org) ha desarrollado un
sistema, puesto al servicio de la comunidad cientfica, en el que combina el mtodo de
Smilek y cols (2005b) con un test comportamental. En un primer momento el
participante elige el color ms adecuado para cada letra y nmero, en tres ocasiones
sucesivas; a continuacin cada letra o nmero es presentado en la pantalla del
ordenador en un color determinado y el participante debe responder, a la mayor
Introduccin _____________________________________________________________
25
brevedad, si el color presentado es el elegido por l en la fase anterior (el color
correcto) o un color distinto (color incorrecto). Aunque una persona, con muy buena
memoria o utilizando distintas estrategias, podra llegar a puntuar de manera similar a
un sinestsico en el test de consistencia, los tiempos de reaccin necesarios para
decidir si el color presentado es el correcto o no distinguen claramente a ambos
grupos.
2.2 Carcter perceptual
Las sensaciones sinestsicas son de carcter perceptual y no basadas en
memoria. En un intervalo de meses, tanto Ramachandran y Hubbard (2001a) como
Smilek, Dixon, Cudahy y Merikle (2001) demostraron, en dos variantes de una tarea de
bsqueda visual, que en sinestsicos que perciben grafemas con colores particulares,
estos colores son percibidos y procesados de una forma muy similar a aquella en la
que los procesaran si estuvieran, de hecho, presentes en el papel o la pantalla del
ordenador.
Ramachandran y Hubbard (2001a) presentaron un conjunto de letras a su
participante sinestsico J C y a un grupo de controles. Entre las letras, presentadas de
manera aleatoria en la pantalla, se esconda una figura geomtrica cuyo contorno
estaba formado por una de las tres letras presentes. Las otras dos se utilizaban como
distractores. Teniendo en cuenta que los participantes saban en cada bloque la
identidad de la letra cuya distribucin formaba la figura geomtrica, la persona con
sinestesia era ms exacta al identificarla cuando el conjunto estimular se presentaba
durante 1s, en comparacin al grupo control.
Smilek y cols. (2001) realizaron una tarea incluso ms exhaustiva en la que
presentaron, durante 32 ms, un grafema enmascarado a su sinestsica C y a un grupo
de controles. La tarea de los participantes era identificar el grafema enmascarado. La
manipulacin crtica consisti en variar el color del fondo en que se presentaban los
estmulos, para que pudiera ser igual o distinto al fotismo evocado por dicho estmulo.
La lgica experimental era que, si los fotismos se perciben en el espacio externo, de
manera automtica y como si fueran colores reales, cuando el color del fondo igualara
al del fotismo evocado por el grafema presentado en gris oscuro, la participante
sinestsica tendra ms dificultad para identificarlo que cuando el fondo no era del
mismo color. Esto fue lo que, de hecho, encontraron. C identificaba los objetivos con
ms facilidad cuando el fondo en el que se presentaban no corresponda con el color
evocado. Los participantes no sinestsicos, no solo no presentaron una diferencia
26
entre ensayos, con fondo congruente e incongruente, sino que tuvieron una ejecucin
general peor que la de C.
En un estudio posterior, adems, demostraron adems que el efecto de
sustitucin de objeto (Enns y Di Lollo, 1997) desapareca para C cuando el fotismo
evocado por el objetivo era distinto al evocado por los distractores. Este efecto ocurre
a nivel conceptual cuando los recursos atencionales estn distribuidos y la mscara
permanece en la pantalla durante un tiempo suficiente tras la presentacin del objetivo
(Di Lollo, Enns y Rensink, 2000). Con esto demostraron que, aunque el efecto de
sustitucin de objeto elimina el reconocimiento consciente de grafemas en los
participantes no sinestsicos, no eliminaba la experiencia de color en C. Aunque la
atencin focalizada parece ser necesaria para el reconocimiento de los grafemas, el
color se suele considerar una caracterstica bsica que no requiere del mecanismo de
binding (Treisman y Gelade, 1980).
2.3 Carcter idiosincrsico
Las percepciones sinestsicas son idiosincrsicas. Aunque los colores que un
sinestsico percibe para un conjunto de letras y nmeros permanece estable a lo largo
del tiempo, su conjunto de colores es totalmente distinto al de otro sinestsico que
tambin vea colores para letras y nmeros. En la Figura 2 se muestran los colores
asociados a letras y nmeros por distintas personas con sinestesia.

Figura 2. Ejemplo de colores para tres personas con sinestesia grafema-color. Como
puede observarse, uno de los casos solo percibe color para las vocales.
Introduccin _____________________________________________________________
27
Los primeros estudios llevados a cabo en la era moderna del estudio de la
sinestesia enfatizaban que los colores asociados con letras o nmeros no seguan un
patrn observable y no estaban determinados por la experiencia previa. As, era muy
normal encontrar la afirmacin de que la sinestesia no estaba causada por juegos de
imanes de la infancia (es comn en algunos pases encontrar juegos para que los
nios aprendan el alfabeto, que consisten en imanes para pegar en el frigorfico con
las letras de distintos colores). Con esto se pretenda enfatizar el hecho de que la
sinestesia no es una asociacin de conceptos sino una experiencia perceptual.
No obstante, algunos estudios, comparando las asociaciones de un gran
nmero de personas con sinestesia, han encontrado tendencias de que ciertos
grafemas estn ms asociados a unos colores que a otros (Simner y cols., 2005,
Simner, Glover y Mowat, 2006). As, por ejemplo, en pases de habla inglesa se ha
encontrado que el color ms probable para la letra A es el rojo, ya que est
fuertemente asociada a manzana (apple), mientras que la B es muy probable que
sea percibida en azul, ya que dicho color empieza por esta letra en ingls (blue).
Estudios recientes han documentado un caso en el que los colores de una persona
con sinestesia estaban en su mayora determinados por un juego de imanes que la
participante tena en su niez (Witthoft y Winawer, 2006).
Una de las participantes sinestsicas estudiada en nuestro laboratorio sostiene
que los colores de sus letras estn determinados por los colores en que se
presentaban en el libro con el que aprendi a leer (Micho). Tras realizar numerosos
estudios experimentales, que se describen a continuacin, conseguimos recuperar un
ejemplar de dicho mtodo de aprendizaje de la lectura. En la Figura 3 se comparan los
colores utilizados en la cartilla con los experimentados por MA.
Puesto que el mtodo de aprendizaje utiliza el criterio fontico para colorear las
letras, puede observarse que letras con el mismo fonema (ej. c y k producen el
fonema /k/) tienen el mismo color y que otras letras (ej. g) tienen distinto color en
funcin del sonido que producen. Tambin es interesante ver que ninguna de las
vocales se presenta en color (excepto la u cuando unida con la g da lugar al sonido
/g/). Por lo tanto, de 27 letras, slo en 11 coincide el color de la cartilla Micho y el
esquema de percepcin sinestsica de MA.

28

Figura 3. Comparacin de los colores utilizados en un mtodo de aprendizaje de lectura
(Cartilla Micho), presentados a la izquierda, con los colores asociados a las letras por la
participante sinestsica MA, presentados a la derecha.
Como se discutir ms adelante, el hecho de que la sinestesia sea un
fenmeno perceptual, automtico e involuntario no est reido con que se conozcan
los orgenes de las asociaciones entre grafemas y colores. Puesto que los humanos
no nacen con conocimiento de los grafemas y sistemas de escritura, es necesario que
tenga lugar el aprendizaje de los mismos para que pueda producirse la sinestesia. Por
lo tanto, aquello que distingue a sinestsicos de los que no lo son sera simplemente la
predisposicin a enlazar informacin aparentemente independiente y experimentarla
de manera vvida e involuntaria. Es posible que los libros utilizados para aprender un
lenguaje o los juegos con los que se interacciona en la infancia sean los determinantes
de las asociaciones especficas de cada persona sinestsica pero no del hecho de que
lo sean.
2.4 Automaticidad
Las percepciones sinestsicas son automticas, involuntarias y difciles de
suprimir. Esto es, la persona que tiene estas experiencias no puede reprimirlas.
Simplemente ocurren. Como antes se mencion, numerosas investigaciones se han
llevado a cabo en este mbito, principalmente con distintas variantes del paradigma
Stroop, para demostrar que las percepciones sinestsicas ocurren de forma
automtica (Dixon y cols., 2002; Elias, Saucier, Hardie y Sarty, 2003; Lupiez y
Callejas, 2006 -Captulo 1-; Mattingley, Rich, Yelland y Bradshaw, 2001; Mills y cols.,
1999, 2002; Odgaard, Flowers y Bradman, 1999; Wollen y Ruggiero, 1983, entre
Introduccin _____________________________________________________________
29
otros). En la versin estndar del paradigma Stroop (1935) se presentan nombres que
designan colores en un color de tinta que puede ser congruente o incongruente con el
color designado por la palabra (ej. ROJ O presentado en color rojo-congruente- o en
color azul-incongruente-). Las respuestas en la condicin incongruente cuando la tarea
del participante es nombrar el color en que se presenta la palabra suelen ser ms
lentas que en la condicin congruente. Como la tarea a realizar no est relacionada
con la lectura de la palabra pero incluso as sta ocurre, se concluye que la lectura de
la palabra es automtica (MacLeod, 1991).
En la extensin del paradigma al estudio de la sinestesia, la manera ms
comn de proceder ha sido la de presentar un conjunto de letras o nmeros en un
color determinado y pedir a los participantes que indiquen el color en que se encuentra
el estmulo. La manipulacin clave, incluida en los estudios de sinestesia consiste en
presentar cada estmulo con el color en el que el sinestsico ha informado
previamente que lo percibe (condicin congruente) o con un color distinto (condicin
incongruente). As una J presentada en rojo podra ser congruente para un
sinestsico mientras que si se presenta en azul sera incongruente. Del mismo modo,
para otra persona con sinestesia ambos colores seran incongruentes y el verde sera
el congruente. En cualquier caso, lo que para una persona sin sinestesia sera una
tarea carente de dificultad (nombrar el color en que se presenta una h o una j), para
una persona con sinestesia debera de ser ms fcil o ms difcil en funcin de la letra
escrita. Es decir, en funcin de que el color evocado por la letra (el fotismo) coincida o
no con el color en el que sta se presenta y por lo tanto, el color que hay que nombrar.
Esto es precisamente lo que se encuentra en una gran variedad de estudios. Las
personas con sinestesia son mucho ms rpidas cuando color real y fotismo coinciden
(condicin congruente) que cuando no lo hacen (condicin incongruente).
Otro enfoque al estudio de la automaticidad de la sinestesia ha sido el de
comparar la interferencia producida por el fotismo con la interferencia producida por el
color real presentado al participante (Dixon y cols., 2004; Lupiez y Callejas, 2006 -
Captulo 1-). Dixon y cols pidieron a diferentes personas con distinto tipo de sinestsia
(asociativa o de proyeccin; ver ms abajo) que realizaran ambas tareas; nombrar el
color y nombrar el fotismo. Encontraron que, en el caso de la sinestesia asociativa, la
interferencia producida por el color real, cuando la tarea consista en nombrar el
fotismo, era mayor que la interferencia producida por el fotismo cuando la tarea era
nombrar el color real del estmulo. Adems, eran ms rpidos nombrando el color real
que nombrando el fotismo. Para la sinestesia de proyeccin encontraron que la
30
interferencia del color real sobre el fotismo era menor y se tardaba menos tiempo en
nombrar el color del fotismo que el color percibido visualmente. Si estos estudios se
generalizan a otros grupos de personas con sinestesia, la tarea de nombrar el color vs.
nombrar el fotismo podra pasar a ser un elemento objetivo para diferenciar ambos
grupos de sinestsicos.
Finalmente, tambin se ha estudiado el grado en que experiencias sinestsicas
se pueden inhibir mediante el paradigma de Priming Negativo (Odgaard y cols, 1999;
Lupiez y Callejas, 2006; Captulo 1).
2.5 Percepcin genrica
Las percepciones sinestsicas son genricas. Es decir, cuando una palabra da
lugar a una percepcin de color, sta es simple, se trata de un color o un pequeo
conjunto de ellos, no de una percepcin elaborada. De igual modo, cuando un
sinestsico, por ejemplo, experimenta formas al probar distintos sabores, estas formas
son muy generales, como lneas o espirales o texturas suaves o rugosas. No
experimentan la sensacin de estar tocando una vaca o viendo un castillo (Cytowic,
1993).
Aunque al describir sus experiencias una persona con sinestesia comente que
el sabor del pollo es como tocar una columna de mrmol, eso es slo una analoga
que hace referencia a un objeto familiar y ayuda en la descripcin. No obstante, si se
insiste en una descripcin basada en sensaciones discretas, las personas con
sinestesia admiten que no es eso lo que perciben sino sensaciones primarias, la
sensacin de ausencia de rugosidad, la temperatura baja, superficie vertical, etc.
2.6 Carcter memorable de las percepciones
Otra caracterstica de la sinestesia es que las percepciones son memorables.
Es comn encontrar que las personas sinestsicas recuerden la percepcin
secundaria o sinestsica de forma ms vvida que la primaria (aquella que evoca la
sensacin sinestsica) (Mills y cols., 2006; Smilek, Dixon, Cudahy y Merikle, 2002).
As, hay casos en los que no recuerdan un nmero de telfono o el nombre de una
persona, pero saben que el nmero empezaba por azul, o que el nombre era verde.
Quizs el caso ms famoso de capacidades de memoria excepcionales sea el
descrito por Luria (1968). S era un experto en mnemotecnia que poda memorizar
Introduccin _____________________________________________________________
31
cincuenta dgitos en tres minutos y recordarlos sin ningn error de manera inmediata y
aos despus. S era, adems, sinestsico y Luria especul que sus habilidades
mnemotcnicas podran ser un resultado de su sinestesia. No obstante, Luria no
proporcion ninguna evidencia emprica de que fuera de hecho la sinestesia lo que
mejoraba su memoria.
Smilek y cols. (2002) realizaron un experimento en el que abordaron el estudio
emprico de esta cuestin. Para ello, presentaron a una participante sinestsica
matrices de nmeros para su memorizacin. Para comprobar que el fotismo evocado
por cada nmero ayudaba a la memorizacin de los mismos, manipularon el color en
que presentaban la matriz a recordar. En un caso (lnea base) la presentaron en negro
y en otras dos ocasiones en color (en un caso congruente con sus fotismos y en otro
incongruente con los mismos). En este estudio encontraron que, cuando los dgitos
estaban presentados en color incongruente, su participante sinestsica C era incapaz
de recordarlos, mientras que, cuando se presentaban en negro o en color congruente,
su recuerdo era mejor que el de la mayora de los controles. Para demostrar que este
efecto no se deba a una buena memoria general realizaron otro experimento en el
que los estmulos a recordar eran caracteres que no evocaban ningn fotismo para C.
En este caso encontraron que su recuerdo no era mejor que el de los controles. Es
ms, en una prueba sorpresa, 48 horas tras la fase de estudio, encontraron que el
recuerdo de los controles para los dgitos y los caracteres decaa drsticamente. Del
mismo modo decaa el recuerdo de C, pero slo para la matriz de caracteres que no
evocaban ningn fotismo. La matriz de dgitos era recordada con la misma exactitud
que en la fase de prueba realizada tras el estudio.
Recientemente Azoulai, Hubbard y Ramachandran (2005) estudiaron el caso
de un joven britnico con capacidades mnemtcnicas excepcionales. En la actualidad
ostenta el record europeo de decimales del nmero pi recordados (22.515). DT es,
adems, sinestsico y sostiene que su sinestesia le ayuda en el recuerdo. Para l,
cada nmero del 0 al 10.000 tiene una forma particular y eso le ayuda a recordarlos y
a realizar operaciones matemticas de manera asombrosamente fcil. Adems, cada
nmero tiene un tamao especfico (6 es el ms pequeo de todos y 9 el mayor).
Azoulai y cols. (2005) presentaron a DT una matriz de 100 nmeros para su
memorizacin. Con Siguiendo una manipulacin similar a la llevada a cabo por Smilek
y cols. (2002), le presentaron a DT una matriz de nmeros en la que el tamao de
cada nmero era igual (lnea base), y un par de matrices en las que cada nmero tena
un tamao. En una matriz los tamaos relativos de los nmeros eran congruentes con
32
lo informado por DT y en otra los tamaos eran incongruentes. Tras cinco minutos de
estudio, DT recordaba el 68% de los dgitos cuando stos aparecan en la condicin
de lnea de base y el 50% cuando aparecan en condicin congruente. Sin embargo,
slo recordaba el 16% cuando aparecan con un tamao incongruente. Lo ms
sorprendente fue que, en un re-test realizado tras un da y en otro realizado tras tres
das, DT recordaba todos los dgitos que haba recuperado en la primera prueba para
las dos primeras condiciones (lnea base y congruente). No obstante, en el caso de los
dgitos con tamao incongruente, a las 24 horas slo record el 4%, y a las 72 horas
no fue capaz de recordar uno solo nmero. Azoulai y colaboradores concluyeron que
las caractersticas sinestsicas asociadas a cada nmero ayudaban a DT a
memorizarlas, pero cuando las caractersticas reales del estmulo interferan con su
percepcin interna, sta ya no era efectiva en la mejora de su capacidad de memoria.
2.7 Unidireccionalidad
Tradicionalmente se ha pensado que la sinestesia es unidireccional (Mills y
cols., 1999) y que un inductor evoca un concurrente; pero, a su vez, dicho concurrente
no evoca el inductor. Esta afirmacin, generalmente aceptada, ha sido cuestionada
recientemente por dos grupos (Brugger y Della Betta, 2000; Brugger, Knoch, Mohr y
Gianotti, 2004; Cohen-Kadosh y cols., 2005; Cohen-Kadosh y Henik, 2006; Knoch,
Gianotti, Mohr y Brugger, 2005;).
Cohen-Kadosh y cols. (2005) utilizaron una variante del paradigma de
congruencia de tamao para estudiar si la sinestesia grafema-color es de hecho
unidireccional. En el paradigma de congruencia de tamao se presenta un par de
nmeros y el participante debe informar sobre el carcter numrico de ellos (ej. indicar
cul es mayor). Si el dgito, que es numricamente mayor, es tambin mayor en
tamao fsico, la tarea se ve facilitada en contraposicin con el caso en el que el dgito,
que es numricamente ms pequeo, es el que aparece en un tamao fsico mayor.
Este resultado se interpreta haciendo referencia al hecho de que el tamao fsico es
procesado automticamente y, por ello, afecta la ejecucin (Henik y Tzelgov, 1982).
Este paradigma de congruencia de tamao se puede modificar para convertirlo
en un paradigma de congruencia de color. No obstante, eso slo es posible si el color
lleva asociado un significado (en este caso, el nmero que lo induce). Para comprobar
si el color evoca el nmero con el que est asociado se presentan nmeros coloreados
con el color que le corresponde o con el color correspondiente a otros nmeros que
estn ms separados ordinalmente o menos separados ordinalmente que los
Introduccin _____________________________________________________________
33
presentados. De esta manera se puede inducir un efecto de congruencia similar pero
slo si el color es efectivo para activar la representacin semntica del nmero con el
que se asocia. Es decir, slo si el concurrente activa a su inductor. Siguiendo esta
lgica Cohen-Kadosh y cols. (2005) presentaron un conjunto de pares de nmeros que
podan tener una distancia numrica pequea (ej. 4 y 5) o una distancia numrica
grande (ej. 2 y 7). En ambos casos los nmeros aparecan en su color correspondiente
(condicin congruente) o en un color asociado con dgitos con distinta distancia
numrica. En el caso del 4 y el 5, podan aparecer con el color del 2 y el 7; y en el
caso del 2 y el 7, podan aparecer con el color del 4 y el 5. Con estas condiciones
encontraron que, cuando el color en que se presentaban los dgitos era el evocado por
dgitos con mayor distancia numrica, los participantes con sinestesia eran ms
rpidos en indicar cul de los dos dgitos era mayor que cuando aparecan en su color
correspondiente. No obstante, el efecto contrario no se encontr (la discriminacin de
dgitos con distancia numrica grande no empeor cuando estaban coloreados con el
color correspondiente a dgitos con pequea distancia numrica). Los autores
concluyeron que, en el primer caso, el procesamiento del color activaba el concepto
numrico correspondiente y ayudaba a la ejecucin. En el ltimo caso, explican la
ausencia de efecto basndose en la diferencia relativa en la velocidad con que se
procesan las dos dimensiones del estmulo. Una distancia numrica grande se
procesara de forma rpida y el procesamiento del color que, a su vez, activara el
concepto numrico no sera llevado a cabo con suficiente rapidez como para influir en
la respuesta.
2.8 Carcter emocional
Por ltimo, la sinestesia tiene un carcter emocional. La experiencia sinestsica
va asociada a una sensacin de certidumbre y conviccin de que lo experimentado es
real y vlido (Cytowic, 1993; 1989/2002). Debido al hecho de que la percepcin
sinestsica ocurre de forma involuntaria y es difcil de suprimir, hace que se
experimente indistintamente de las caractersticas del inductor. As, una J que lleva
asociada la experiencia del color verde, seguir induciendo dicho color
independientemente de que est escrita en negro, verde o amarillo. Esta presencia o
ausencia de coincidencia entre las caractersticas preceptuales del inductor y del
concurrente experimentado lleva a crear situaciones de congruencia o incongruencia
entre ambas dimensiones, que inducen ciertas reacciones emocionales en las
personas sinestsicas, bien de carcter positivo o negativo. Muchas personas con
sinestesia informan de que la percepcin de algo inconsistente, con su experiencia
34
subjetiva, es desagradable y les causa malestar. Ramachandran y Hubbard (2001b)
comentaron brevemente este aspecto de la sinestesia e incluso especularon con la
posibilidad de que la presencia de conexiones cruzadas en el cerebro de las personas
con sinestesia no se viera reducida a la zona que procesa la forma y el color sino que
estuviera presente en toda la arquitectura cerebral, incluidas zonas del sistema
lmbico. Esto explicara la experiencia de dichas reacciones afectivas. Esta
caracterstica de la sinestesia es una de las menos estudiadas hasta el momento y es,
precisamente, en la que se centra el presente trabajo de investigacin.
3 Prevalencia y carcter hereditario
Ya en el siglo XIX, Galton (1883) advirti que el tipo de sinestesia ms comn
era la escucha coloreada y que era mucho ms comn entre miembros de una misma
familia de lo que lo era en la poblacin general. Poco tiempo ms tarde, en 1893
Calkins estim a partir de una muestra de ms de 500 personas, que el 6.7% de los
entrevistados experimentaban sinestesia (entonces la llam pseudo cromatoestesia).
De este grupo, slo el 1% inform percibir color para todas las palabras. Baron-Cohen
y colaboradores (1996) realizaron un estudio que ha influido enormemente en la
investigacin posterior, tanto por sus afirmaciones sobre la prevalencia de la
sinestesia, como sobre el posible componente hereditario de la misma y el porcentaje
de hombres y mujeres que la experimentan. Tras anunciar su investigacin en un
peridico local, un grupo de personas con sinestesia se pusieron en contacto con su
laboratorio. Basndose en las estadsticas del nmero potencial de lectores y en el
nmero de personas que se pusieron en contacto con ellos para comunicar su
sinestesia, llegaron a la estimacin de que 1 de cada 2000 personas experimenta
sinestesia (0.05%). Tambin encontraron una proporcin mayor de mujeres que de
hombres (de 6:1 a 3:1 en funcin de que el estudio se realizara con habitantes de la
ciudad de Cambridge o con estudiantes de dicha universidad). En ambos casos
encontraron adems, al igual que Galton, que en torno al 25-30% de ellos informaban
tener alguien en su familia que tambin experimentaba sinestesia.
Basndose en este estudio, Baron-Cohen y cols. (1996; Bailey y J onson, 1997)
concluyeron que el patrn de transmisin gentica, que ms se ajustaba a tales
caractersticas, era la transmisin ligada al sexo con letalidad. Propusieron que un solo
gen ligado al cromosoma X podra ser el responsable del patrn encontrado y que la
mayor cantidad de mujeres que de hombres podra deberse a la concurrencia de
Introduccin _____________________________________________________________
35
letalidad. Es decir, de los varones engendrados y portadores del gen de la sinestesia,
la mitad no llegaran a trmino (daran lugar a abortos espontneos).
La prediccin clara de esta propuesta es que las madres con sinestesia
tendran ms hijas que hijos y un mayor nmero de abortos espontneos que la
poblacin general. Aunque la primera afirmacin pareca estar apoyada por los datos
de Baron-Cohen y cols. (1996), la segunda no ha sido comprobada hasta
recientemente. Ward y Simner (2005), recientemente, realizaron un estudio con 85
familias y concluyeron que, aunque sus resultados eran bsicamente consistentes con
la hiptesis de una herencia ligada al cromosoma X, en su muestra no haba evidencia
de letalidad en los varones, ya que encontraron un nmero de hijos similar al de hijas
de personas con sinestesia.
Un par de artculos recientes plantean cierto conflicto con el punto de vista de
que la sinestesia se puede explicar con un patrn de herencia dominante ligada al
cromosoma X. Smilek y cols. (2002) documentaron un caso de una pareja de gemelas
monocigticas discordantes para sinestesia. Un anlisis genotpico demostr que su
carga gentica era idntica. No obstante, pruebas objetivas, as como sus propios
informes subjetivos, demostraron que slo una de ellas era sinestsica. Los autores
propusieron como posible explicacin de dicha discordancia el hecho de que en cada
una de las gemelas se hubiera desactivado un cromosoma X distinto. No obstante,
recientemente, informaron de un caso que es ms difcil de conciliar con el punto de
vista mayoritario sobre el modo de herencia de la sinestesia. (Smilek, Dixon y Merikle
(2005b). Han documentado un caso en el que dos hermanos gemelos monocigticos
son discordantes para la sinestesia. Puesto que en los varones slo hay un cromosma
X, la hiptesis de la desactivacin no puede explicar este caso. Aunque una posible
explicacin hace referencia a la relativa penetracin del gen, los autores se decantan
por cuestionar el hecho de que la sinestesia sea hereditaria. Basndose en estudios
recientes, que demuestran una mayor prevalencia de la sinestesia que lo
anteriormente pensado, (Stephan, 2005; Simner y cols., en prensa), los autores
cuestionan que, realmente, sea ms comn encontrar a distintas personas con
sinestesia dentro de una familia que en la poblacin general.
Tres laboratorios estn intentando descifrar el posible gen responsable de la
sinestesia y su modo de actuacin (Asher, Lamb, Baron-Cohen y Monaco en
Cambridge, Reino Unido; Eagleman, Kagan, Sarma y Nelson en Houston, Estados
Unidos y Mitchell, Bargary, Barnett y Newell en Irlanda). El primer grupo se decanta
por la idea de que la sinestesia est ligada a ms de un cromosoma (Asher, Lamb,
36
Baron-Cohen y Monaco, 2006) y no predice un mecanismo de accin particular de los
mismos. El segundo grupo se decanta por la posibilidad de que la accin del gen
ligado a la sinestesia sea la de producir una arquitectura cerebral distinta de la de
personas no sinestsicas (Eagleman, Kagan, Sarma y Nelson, 2006; ver hiptesis de
la activacin cruzada ms abajo). Por ltimo el tercer grupo se centra en la posibilidad
de que el gen ligado a la sinestesia tenga implicaciones funcionales (Mitchell, Bargary,
Barnett y Newell, 2006; ver hiptesis de la retroalimentacin desinhibida ms abajo).
Igualmente, en lo referente al sesgo encontrado entre el nmero de mujeres y
hombres con sinestesia, existe un cierto debate sobre la exactitud y la interpretacin
de estas afirmaciones. Aunque estudios recientes encuentran unas estimaciones muy
similares (Rich y cols., 2005), es importante tener en cuenta que ambos estudios
(Baron-Cohen y cols., 1996 y Rich y cols., 2005) se basaron en muestras auto-
referidas. Una informacin publicada en un peridico local, en un caso, y nacional en
el otro, sirvieron de reclamo para que todo aquel interesado contactara con los
investigadores. Las caractersticas de este mtodo tienen inherente un problema
relacionado con la propensin de distintos subtipos de poblacin a responder ms que
otros. En particular, la diferencia encontrada entre el nmero de hombres y mujeres,
que contestaron al reclamo, podra deberse al hecho de que los hombres sean menos
propensos a ofrecer informacin sobre sus peculiaridades que las mujeres, y no tanto
a una mayor prevalencia de la sinestesia entre las mujeres. Rich y colaboradores
(2005) mantienen que el hecho de que la proporcin de personas con sinestesia en
Australia sea 6.1:1 (extremadamente parecida al 6.3:1 de Baron-Cohen y cols, 1996,
en Cambridge, Reino Unido) no puede explicarse slo sobre la base de un sesgo de
gnero en la propensin a ofrecer informacin de este tipo. Day (2005) encuentra una
proporcin similar con los datos obtenidos mayoritariamente con poblacin
estadounidense.
No obstante, las razones encontradas en otros estudios difieren
significativamente de estas. Recientemente Simner y cols. (en prensa) realizaron dos
estudios a gran escala en los que las muestras no eran auto-referidas. En ambos
casos se realizaron pruebas objetivas y subjetivas para determinar la ocurrencia de la
sinestesia grafema-color o de cualquier otro tipo de sinestesia. En el primer estudio
realizaron una pequea prueba a ms de mil visitantes del Museo de la Ciencia de
Londres, en una exposicin interactiva. A cada participante le presentaron todas las
letras y nmeros de manera individual y se le pidi que le asignaran el color que mejor
combinara con dicho estmulo. Una vez terminada la prueba se le volvi a presentar el
Introduccin _____________________________________________________________
37
mismo conjunto de estmulos y se infiri de ambas puntuaciones una medida de
consistencia. Por ltimo preguntaron a los visitantes si pensaban que ellos
experimentaban sinestesia. De los 1190 visitantes, que realizaron las pruebas, el 1.1%
fue identificado con sinestesia grafema-color. Una prueba similar se llev a cabo con
500 estudiantes universitarios escoceses y, de nuevo, el porcentaje de personas con
sinestesia grafema-color era del 1.4%. An ms interesante fue el hecho de que el
mismo nmero de hombres que de mujeres fueron identificados en estas muestras.
Estos datos suponen el primer intento de estudio de la prevalencia de la sinestesia en
una muestra ms aleatoria de la poblacin. En este sentido, es importante sealar
que, aun cuando en este estudio el participante no era el agente activo que deba
informar de su percepcin, la muestra tambin estaba sesgada en cierta manera, pues
slo determinados grupos de poblacin estn interesados en visitar un museo de
ciencia y, obviamente, la muestra universitaria es slo representativa de ese subgrupo
de la poblacin). Datos similares, en cuanto a la prevalencia de la sinestesia,
provienen del estudio de Mulvenna, Hubbard, Ramachandran y Pollick (2004).
4 Tipos de sinestesia
Ya en 2001 Ramachandran y Hubbard propusieron que podra haber dos
subgrupos de gente con sinestesia, dentro de la sinestesia grafema-color. Los
denominaron sinestsicos de alto nivel y de bajo nivel y lanzaron la hiptesis de
que el locus de la activacin cruzada estara localizado en diferentes reas
cerebrales en ambos grupos (ver teora de la activacin cruzada ms abajo). Ambos
grupos se diferenciaran, en la prctica, por el nivel de procesamiento del inductor, en
el que se producira la activacin del concurrente. Mientras que en los sinestsicos de
bajo nivel la forma del grafema sera el desencadenante del color, en los de alto
nivel seran las caractersticas ordinales de los estmulos (el orden de las letras en el
abecedario o de los nmeros o das de la semana). El inductor respondera, por lo
tanto, a un nivel ms conceptual del mismo estmulo.
De forma casi paralela Smilek y Dixon (2002) propusieron una distincin en
funcin de los informes subjetivos de las personas sinestsicas con las que haban
interaccionado. Al describir sus experiencias, un gran nmero de sinestsicos hacan
referencia al lugar espacial en el que perciban los fotismos. Mientras unos informaban
que el color apareca proyectado en el espacio, normalmente encima del estmulo que
lo induca, otros aseguraban que el color no lo vean proyectado, sino en la mente.
Estos informes llevaron a los autores a proponer dos tipos de sinestesia grafema-color,
38
la sinestesia de proyeccin y la sinestesia de asociacin. Postularon que la primera
estara caracterizada por una mayor automaticidad en la percepcin del fotismo que la
segunda.
No obstante, algunos investigadores (Edquist, Rich, Brinkman y Mattingley,
2006; Mattingley y cols., 2001; Rich y Mattingley, 2005) obvian la posibilidad de que
haya diferentes subgrupos de sinestesia grafema-color y agrupan a personas con
sinestesia en sus anlisis sin evaluar la posibilidad de que sus experiencias se puedan
catalogar bien en uno de los grupos descritos por Ramachandran y Hubbard (2001b),
bien o en uno de los descritos por Dixon y Smilek (2005).
Recientemente Dixon y cols. (2004) han demostrado empricamente que las
personas con sinestesia de proyeccin tienen un patrn de comportamiento distinto a
las personas con sinestesia de asociacin. Al tomar un grupo de personas, que
explicaban sus experiencias sinestsicas haciendo referencia a que el color apareca
en la mente, y otro grupo, que haca referencia al color como proyectado en el
espacio, realizaron un experimento Stroop con dos variantes. En un caso, pedan a los
participantes nombrar el color en que apareca el estmulo e ignorar su fotismo y, en
otro, pedan nombrar el fotismo e ignorar el color presente en la pantalla del
ordenador. La interferencia causada por el color real al nombrar el fotismo fue mayor
en el grupo de sinestesia de asociacin, mientras que la interferencia causada por el
fotismo al nombrar el color real fue mayor en el grupo de sinestesia de proyeccin. Los
autores concluyeron que la sinestesia de proyeccin tena un carcter ms perceptual
que la de asociacin y, por ello, los ltimos tenan ms interferencia del color real al
nombrar el fotismo experimentado en la mente.
Si bien es cierto que muchas personas con sinestesia no se ven identificadas
con ninguna de estas dos categoras, eso no es motivo suficiente para obviarlas, ya
que dan lugar a patrones comportamentales distintos. En lugar de obviar las
diferencias, sera conveniente que seguir profundizando en las caractersticas de este
tipo de sinestesia, ya que puede haber otros tipos intermedios o simplemente tratarse
de un continuo de viveza con la que se experimenta el fotismo. En cualquier caso, la
diferenciacin de los diferentes subtipos de sinestesia es importante, ya que nos
permitir un mejor conocimiento de la sinestesia y probablemente de los diferentes
mecanismos cognitivos y cerebrales que la sustentan.
Introduccin _____________________________________________________________
39
5 Bases cerebrales
El nmero de estudios con imagen cerebral llevados a cabo en la investigacin
de las bases neuronales de la sinestesia va en aumento, y de forma especial, en los
ltimos aos. En un primer estudio, Paulesu y cols. (1995) utilizaron PET para estudiar
la sinestesia palabra-color. Compararon la actividad cerebral de personas con
sinestesia y participantes control al escuchar tonos puros o palabras habladas. Las
personas sinestsicas que estudiaron perciban colores para el lenguaje hablado pero
no para los tonos puros. Comparando la activacin entre en ambas condiciones, en
personas con sinestesia encontraron activacin en reas del cortex temporal posterior-
inferior y en zonas de la unin parieto-occipital, pero ninguna diferencia en zonas de
procesamiento visual primario como V1, V2 o V4. Ninguna de estas zonas estaba
activa en controles no sinestsicos.
En un estudio posterior Nunn y cols. (2002) estudiaron de nuevo sinestsicos,
para los que la escucha de nombres evocaba color, pero con tcnicas de resonancia
magntica funcional. En este caso encontraron activacin diferencial en la zona del
cerebro relacionada con el procesamiento del color (V4/V8), pero no en zonas de
procesamiento visual ms tempranas (V1 o V2,) siempre que las personas con
sinestesia escuchaban nombres pero no cuando escuchaban sonidos puros. De nuevo
ninguna de estas activaciones estaba presente en los participantes del grupo control.
Otros estudios posteriores de caso nico tambin han encontrado patrones de
actividad inconsistentes. Mientras que Weiss y cols. (2001) encontraron activacin
cerca de V4, pero no en V1, en una persona con sinestesia, para la que los nombres
de persona evocaban colores, en un estudio ms reciente no encontraron actividad en
estas zonas cerebrales sino en zonas del surco intraparietal (Weiss, Zilles y Fink,
2005) que se han relacionado con la unin de color y forma (Robertson, 2003). Por
ltimo Aleman y cols. (2001) encontraron activacin en V1 pero no en V4. Segn
Hubbard y Ramachandran (2005) esta discrepancia en cuanto a las zonas activadas
podra deberse a diferencias metodolgicas o al tipo de sinestesia de las personas
estudiadas.
Para evaluar la contribucin relativa de estas dos posibles variables
contaminantes, Hubbard y cols. (2005) combinaron un conjunto de pruebas
psicofsicas con un estudio de imagen cerebral y un par de tareas secundarias, a fin de
determinar anatmicamente las reas visuales en cada uno de los participantes
sinestsicos y sus respectivos controles. Mediante la utilizacin de tcnicas de
40
estimulacin de la retina como cuas blancas y negras en rotacin o anillos
expandindose (Engel, Rumelharth, Wandell y Lee, 1994; Sereno y cols., 1995),
identificaron las reas visuales primarias de cada participante que estn organizadas
de forma retinotpica con un margen de error de 2 a 4mm (Engel, Zhang y Wandell,
1997). Hubbard y cols. (2005) encontraron que V4 (pero no reas anteriores en el
procesamiento visual: V1, V2) estaba ms activado, en personas con sinestesia,
cuando perciban grafemas que evocaban color que cuando perciban caracteres
grficos sin significado (y por lo tanto sin color). Es interesante comentar que la
activacin encontrada en estas zonas correlacionaba con la mejora encontrada en las
tareas psicofsicas en relacin con los controles. Por lo tanto, parece ser que las
diferencias individuales en ejecucin entre distintas personas con sinestesia van
acompaadas de cambios cuantitativos en la actividad cerebral de las reas de
procesamiento del concurrente.
Otro estudio reciente tambin compar la actividad de personas sinestsicas
con controles delimitando las zonas visuales para cada uno de los participantes
(Sperling y cols., 2006). De nuevo encontraron que la activacin de V4 era mayor en
sinestsicos cuando se le presentaban grafemas para los que tenan una experiencia
sinestsica de color que para aquellos que no la producan.
Por lo tanto, los estudios ms recientes, que realizan un mapa retinotpico para
cada uno de los participantes, parecen coincidir en el hecho de que V4 est ms
activada cuando personas con sinestesia experimentan un color que cuando no lo
hacen.
Aunque estos estudios informan de actividad en V4, la baja resolucin temporal
de las tcnicas de imagen cerebral no ofrece informacin sobre el momento temporal
en que dicha activacin se produce y, por lo tanto, no ayudan a discernir si tal actividad
es concurrente a la del procesamiento de la forma o posterior. Estudios de
electrofisiologa son necesarios para discernir este punto.
En este sentido slo un estudio se ha llevado a cabo hasta el momento sobre el
curso temporal de las percepciones sinestsicas (Schiltz y cols., 1999). Un conjunto de
17 personas con sinestesia y otros tantos controles realizaron una tarea, tpica del
paradigma de odd-ball, en la que haba que detectar un objetivo infrecuente
entremezclado con distractores. Ambos, objetivo y distractor, podan ser letras o
nmeros. Los resultados del estudio demostraron que el grupo de sinestsicos
comenzaba a diferenciarse de los controles alrededor de los 150ms post aparicin del
Introduccin _____________________________________________________________
41
objetivo y en electrodos frontales y centrales, pero no en los parietales. Los autores
interpretaron los datos haciendo referencia a las demandas de la tarea en la que las
personas con sinestesia tenan que ignorar el color evocado por el estmulo para
centrarse en su identidad y as poder realizar la tarea correctamente, tal y como
haban propuesto anteriormente Paulesu y cols. (1995) para explicar cierta actividad
encontrada en su estudio de PET en reas prefrontales derechas.
No cabe duda de que son necesarios ms estudios electrofisiolgicos para
llegar a un mejor entendimiento del curso temporal de las experiencias subjetivas
asociadas a la sinestesia.
6 Teoras Explicativas
Han sido numerosas las teoras propuestas para explicar el fenmeno de la
sinestesia, tanto desde un punto de vista psicolgico como neurocognitivo. Tambin
han sido mltiples los modos de clasificar dichas teoras, en funcin del nfasis dado a
ciertas caractersticas de unas teoras frente a otras (Harrison y Baron-Cohen, 1997;
Harrison, 2001; Cytowic, 2002; Marks y Odgaard, 2005; Hubbard y Ramachandran,
2005). La estructura con la que se enuncian a continuacin utiliza como criterio el
mbito en que se plantea la teora, ya sea psicolgico o neurocognitivo y la relacin
que postulan entre las personas no sinestsicas y las sinestsicas. Como podr
observarse, las teoras estrictamente psicolgicas pertenecen a los estudios clsicos
de principios del s.XIX y las neuropsicolgicas a los estudios actuales.
6.1 Teoras Psicolgicas
Dos teoras se pueden diferenciar en este campo. Ambas hacen referencia a
asociaciones creadas entre el inductor y el concurrente que explican la experiencia
subjetiva de uno al percibir el otro. As la sinestesia sera el producto de una cadena
de asociaciones mentales en la que algunos de los pasos intermedios podran haber
dejado de ser conscientes. La diferencia entre ambas teoras radica en el mecanismo
postulado para la aparicin de dichas asociaciones.
La Teora de las asociaciones (Langfield, 1926: Wheeler y Cutsforth, 1922)
explica la sinestesia haciendo referencia a asociaciones azarosas de manera que si A
sugiere B, entonces A y B deben haber sido experimentadas de manera simultnea en
algn momento del pasado. Esta teora tiene problemas para explicar el hecho de que,
42
en la misma familia, distintos miembros experimenten asociaciones distintas entre, por
ejemplo, letras y colores.
La Teora del tono emocional (Calkins, 1895, Smith, 1905) propone una
explicacin similar, pero haciendo referencia al significado connotativo de los estmulos
implicados. Propone que el inductor y el concurrente comparten un trasfondo
emocional que es el causante de la asociacin entre ellos. Tal y como Cytowic (2002)
plantea, esto llevara a postular que un sonido estara, no slo asociado a un color con
el que comparte la misma base emocional, sino tambin a un sabor, un olor, etc.
En la actualidad estas teoras han recibido poco apoyo ya que los estudios
comportamentales y de neuroimagen parecen dejar patente que los mecanismos
implicados en la sinestesia tienen una base neural. En cualquier caso, el hecho de que
las nuevas teoras neurocognitivas puedan explicar de un modo ms preciso los
mecanismos responsables de la sinestesia en s, teoras psicolgicas parecidas a
estas podran ser adecuadas para explicar la gnesis de las asociaciones especficas
de cada sinestsico.
6.2 Teoras Neurocognitivas
Desde la Neurociencia Cognitiva, el rea de conocimiento ms dedicada en la
actualidad al estudio de la sinestesia, se han propuesto diferentes teoras en las que
se combinan mecanismos cognitivos y cerebrales para la explicacin del fenmeno.
Estas teoras recogen los datos sobre reas cerebrales implicadas en la sinestesia
provenientes de distintas tcnicas de neuroimagen (ver seccin anterior) y plantean
distintos modos en los que las zonas encargadas del procesamiento del concurrente
se veran activadas cuando se procesa el inductor, dando as lugar a la experiencia
subjetiva de sinestesia. Adems, tambin plantean si los mecanismos que hacen
posible esta comunicacin estn presentes en todos los humanos y, por alguna razn,
slo activos en las personas con sinestesia o si no son compartidos por todos. Aunque
algunos autores (Hubbard y Ramachandran, 2005) plantean que estas dos
dimensiones son independientes, no todas las posibilidades han sido propuestas. La
mayor parte de las teoras se han basado en estudios de sinestesia grafema-color
aunque tambin hay alguna que se centra en otros tipos de sinestesia como la
sinestesia gustativo-tctil (Cytowic, 1993).
Introduccin _____________________________________________________________
43
Teora de la activacin cruzada local (Ramachandran y Hubbard)
La hiptesis de que la sinestesia pudiera deberse a una activacin cruzada
entre reas encargadas del procesamiento de la forma y procesamiento del color fue
propuesta hace ms de un siglo y retomada posteriormente por otros investigadores
(Harrison y Baron-Cohen, 1997; Marks, 1975; Grossenbacher, 1997). No obstante, no
ha sido hasta recientemente cuando se ha presentado de modo elaborado y
comprensible. Ramachandran y Hubbard (2001) la retomaron y ampliaron para
proponer un locus especfico de dicha activacin cruzada. Basndose en estudios
previos sobre el rea de la Forma Visual de las Palabras (Cohen y Dehaene, 2004) y
el homlogo humano a V4, que procesa el color (Wade y otros, 2002), propusieron que
en esta zona de la corteza fusiforme podra estar producindose la activacin cruzada
que dara lugar a la experiencia subjetiva de color, cuando se procesara visualmente
un grafema. Adems, para dar cuenta de las diferencias entre distintos tipos de
sinestsicos, aadieron que la activacin cruzada tambin podra ocurrir en niveles
superiores del procesamiento de la forma y del color, en el giro angular izquierdo. Esta
zona se ha implicado en los clculos numricos abstractos (Dehaene, 1997). Cerca del
giro angular est el giro temporal superior que se ha relacionado con el siguiente
estado en el procesamiento del color (Zeki y Marini, 1998). Por lo tanto, la teora
propone que al percibir un grafema, la informacin llega a zonas de la corteza estriada
y extraestriada y posteriormente es enviada a la zona del rea de la forma visual de la
palabra y de ah a la corteza angular. En uno de estos pasos se produce una
activacin cruzada de zonas cerebrales adyacentes que procesan el color y dicha
activacin da lugar a la experiencia subjetiva de color que acompaa a la experiencia
del estmulo visual que se est percibiendo. En la Figura 4 se esquematiza esta teora
y los posibles estados del procesamiento en los que se producira la activacin
cruzada entre zonas adyacentes. Esta activacin cruzada lleva a la experiencia
reproducible, involuntaria y sistemtica de percepciones sinestsicas (Hubbard y
Ramachandran, 2003). Aunque la teora fue propuesta dentro del marco de la
sinestesia grafema-color, los autores proponen que podra extenderse a otros tipos de
sinestesia en los que las zonas cerebrales encargadas de procesar el inductor y el
concurrente se encuentran prximas dentro de la geografa cerebral.
Respecto al origen de estas conexiones cruzadas entre zonas adyacentes del
cerebro, Ramachandran y Hubbard (2001b; Hubbard y Ramachandran, 2005)
proponen que son debidas a una deficiente poda neural en las etapas tempranas del
desarrollo cerebral. Aunque no descartan la posibilidad de que dichas conexiones
44
estn presentes en la poblacin adulta y slo sean funcionales en las personas con
sinestesia (tal vez por encontrarse desinhibidos), se decantan por la posibilidad de que
una mutacin gentica d lugar a un exceso de conexiones que en condiciones
normales no se vera. Esta hiptesis la apoyan en el hecho de que en fetos de macaco
existe un 70%-90% de conexiones de reas superiores con V4, mientras que las
mismas conexiones no superan el 20%-30% en el macaco adulto no superan el 20%-
30%. (Kennedy, Batardiere, Dehay y Barone, 1997). Si a causa de una mutacin
gentica estas conexiones continan presentes en el adulto, podran llevar a la
experiencia de color cuando se visualizan letras o nmeros. Por lo tanto, aunque las
zonas encargadas del procesamiento del color y de la forma estn adyacentes, el
mecanismo determinante de que exista la activacin cruzada no sera esa proximidad
espacial sino el hecho de que las conexiones existentes en el perodo perinatal no se
poden. En trabajos ms recientes Hubbard y Ramachandran (2005) admiten la idea de
que el mecanismo causante de la sinestesia pueda ser comn a todos los adultos pero
est presente, en mayor grado, en las personas con sinestesia. Ese mecanismo
compartido sera el causante de las correspondencias entre algunos mapas cross-
sensoriales, como la intensidad luminosa y la sonora (Ward y cols., 2006) o entre las
figuras onduladas o puntiagudas y sonidos como bouba o kiki (Ramachandran y
Hubbard, 2001b).

Figura 4. Representacin esquemtica de la teora de la activacin cruzada de
Ramachandran y Hubbard, 2001
Introduccin _____________________________________________________________
45
Esta teora recibe un fuerte apoyo experimental de las investigaciones de
Maurer (1997, Maurer y Mondloch, 2005) que propone que todos los humanos somos
sinestsicos durante los dos o tres primeros meses de vida. En sus estudios con
recin nacidos encontr que bebs habituados a una intensidad luminosa particular no
respondan de modo especial a un estmulo sonoro de igual intensidad pero s a uno
de intensidad diferente. Maurer propone dos variantes de la hiptesis para explicar sus
resultados. La hiptesis fuerte propone que los recin nacidos se parecen a los adultos
sinestsicos, para los que la estimulacin de una modalidad sensorial evoca el
precepto de esa modalidad, y, adems un precepto en una modalidad secundaria. Por
lo tanto el recin nacido tendra percepciones sinestsicas porque algunas de las
conexiones transitorias estaran activas. En la forma dbil de su hiptesis (Maurer y
Mondloch, 1996) propone que los bebs no diferencian estmulos de distintas
modalidades a causa de la inmadurez de su corteza cerebral. Por lo tanto, estaran
respondiendo a la cantidad total de energa sumada a travs de todas las modalidades
sensoriales. Esto implica que seran conscientes de los cambios en el patrn de
energa y capaces de reconocer algunos patrones experimentados con anterioridad,
pero seran incapaces de diferenciar qu modalidad fue la que produjo el patrn de
energa.
Teora de retroalimentacin desinhibida (Grossenbacher y Lovelace)
Esta teora se basa en el hecho de que las conexiones que unen regiones
sensoriales con zonas superiores de procesamiento (conexiones de abajo hacia
arriba) se ven completadas con conexiones de retroalimentacin (de arriba hacia
abajo) que informan a regiones inferiores de procesamiento sensorial de lo procesado
en estados posteriores (Cynader y cols., 1988). Estas conexiones de
retroalimentacin pueden ser de suma importancia cuando la percepcin de un
estmulo sensorial va precedida por una expectacin del mismo (modulacin de arriba
hacia abajo).
Grossenbacher (1997, Grossenbacher y Lovelace, 2001) proponen que las
conexiones de retroalimentacin son una alternativa a los modelos de activacin
cruzada. Segn este modelo el procesamiento del inductor es llevado a cabo de modo
rutinario y, en un momento dado, en un nexo multimodal, la informacin procedente del
procesamiento del inductor activara vas de retroalimentacin en la ruta normal de
procesamiento del concurrente, dando como resultado el inicio de un procesamiento
del concurrente sin estimulacin directa. Por lo tanto, las seales hacia delante en la
46
va de reprocesamiento del inductor activaran neuronas en una zona cerebral donde
la va de procesamiento de ste y del concurrente convergen. Las seales de
retroalimentacin de este rea se propagaran hacia abajo en la va de procesamiento
del concurrente, activando as su representacin. En la Figura 5 se esquematiza esta
teora.
En la mayora de las personas estas vas de retroalimentacin de arriba hacia
abajo estn inhibidas y por lo tanto no experimentan induccin sinesttica. En cambio,
en las personas con sinestesia estas vas de retroalimentacin estaran desinhibidas y
la informacin que entra en las zonas de convergencia se propaga por la va de
procesamiento del concurrente. Por lo tanto, esta teora propone que la arquitectura
cerebral de sinestsicos y no sinestsicos es equivalente y la diferencia en percepcin
subjetiva de ambos grupos se debera a la desinhibicin de algunas de las vas
compartidas por todos.

Figura 5. Representacin esquemtica de la teora de la retroalimentacin desinhibida de
Grossenbacher y Lovelace (2001).
Esta teora tambin es defendida por Cytowic (2002). En un principio propuso
la mediacin de estructuras lmbicas en la sinestesia y la supresin de actividad
cortical (Cytowic, 1993). Dicha hiptesis estaba basada en estudios de actividad
cerebral llevados a cabo con xenon-133 en los que encontr que una persona con
Introduccin _____________________________________________________________
47
sinestesia, en la que sabores eran experimentados con cierta forma, la actividad
cerebral descenda hasta niveles asombrosos cuando experimentaba sinestesia. Esto,
aadido al hecho de que Cytowic considera de suma importancia el afecto asociado a
la experiencia sinestsica, as como la capacidad memorstica aumentada, lo llev a
concluir que unas partes del cerebro se desconectan de otras haciendo que los
procesos normales del sistema lmbico se disparen y salgan a conciencia, siendo
experimentados como sinestesia. No obstante, en publicaciones ms recientes
(Cytowic, 2002) admite que la retroalimentacin de arriba hacia abajo desde zonas
transmodales a zonas unimodales explica de manera simple tanto el carcter
memorable como el afecto asociado a la sinestesia, ya que uno de los enlaces
transmodales es el complejo hipocampal-entorinal. Cytowic favorece esta teora en
detrimento de la teora de la activacin cruzada local ya que con los parmetros de
este enfoque se puede explicar cualquier tipo de sinestesia y las sensaciones
asociadas a sta y no solo la sinestesia grafema-color.
Una variante de esta teora fue propuesta por Smilek (Smilek, Dixon, Cudahy y
Merikle, 2001) especficamente para explicar la sinestesia grafema-color. La
denominaron teora del procesamiento reentrante y propusieron que la sinestesia
grafema-color est causada por retroalimentacin desde zonas del giro fusiforme
anterior, donde se accede al significado de los dgitos (Allison y cols, 1994). La
retroalimentacin de estas zonas hacia las zonas que procesan el color (hV4) hara
cambiar la percepcin del dgito presentado externamente. Segn esta hiptesis, el
nexo multimodal, en el que se llevara a cabo el proceso de retroalimentacin a otras
zonas, estara en el giro fusiforme anterior La Figura 6 muestra una comparativa de
esta teora explicativa de la sinestesia grafema-color y de la teora de la activacin
cruzada local (Ramachandran y Hubbard, 2001b).
Una serie de estudios posteriores (Myles, Dixon, Smilek y Merikle, 2003: Dixon
y cols., 2004; Smilek, Dixon y Merikle, 2005a; Dixon, Smilek, Duffy, Zanna y Merikle,
2006) han mostrado que, incluso en personas con sinestesia de muy bajo nivel, una
misma forma fsica puede evocar un color u otro en funcin del significado que se le
asigne. As, un estmulo ambiguo, que puede identificarse como una letra S o el
nmero 5, puede evocar el color asociado con la letra o el asociado con el nmero, en
funcin del contexto en que se presenta. Estos autores mantienen que las propuestas
de activacin cruzada, entre zonas que procesan la forma y el color, no pueden
explicar estos resultados y, por lo tanto, postulan que es el significado y no la forma
fsica el que evoca el fotismo experimentado.
48

Figura 6. Representacin esquemtica de la teora del procesamiento reentrante de
Smilek, Dixon y Merikle (2001) y comparacin con la teora de activacin cruzada de
Ramachandran y Hubbard (2001b).
No est claro, a partir de lo presentado por los autores, si las vas de
retroalimentacin postuladas estaran presentes en todas las personas, y slo activas
en los sinestsicos, o si seran causa de una arquitectura cerebral diferente. No
obstante, el uso de trminos como aberrante o anormal para referirse a las conexiones
de retroalimentacin hace pensar que los autores se decantan ms por una base
neural diferenciada en las personas con sinestesia.

Los datos de los que actualmente se dispone no son suficientemente finos
como para poder distinguir entre una teora y otra. Smilek et al. (2005a; Dixon et al.,
2006) sostienen que los datos provenientes de experimentos en los que una misma
forma evoca distintos grafemas en funcin del contexto en que se encuentre son
evidencia a favor de que la forma, junto con el significado son necesarios para la
experiencia sinestsica. A su vez, esta modulacin contextual de la sinestesia la toman
como evidencia a favor de su teora ya que la informacin debe ser procesada hasta
Introduccin _____________________________________________________________
49
niveles de significado antes de que pueda experimentarse el color sinestsico. No
obstante, Hubbard y Ramachandran (2005) argumentan que las conexiones de
retroalimentacin normales existentes entre distintas reas cerebrales son suficientes
para explicar los efectos de contexto desde su teora. En la Figura 6 se ve cmo ellos
proponen que la informacin procedente de las zonas de procesamiento del
significado llegara a V4 a travs de el rea de Anlisis de la forma en vez de
directamente como proponen Smilek et al. (2001).
Es probable que los resultados de estudios genticos o de las nuevas tcnicas
de imagen cerebral de conexiones entre reas sean capaces de discriminar entre una
teora u otra. Los primeros informarn de si realmente la sinestesia tiene un
componente gentico y, si as es, de la funcin del gen o genes relacionados con la
misma. Los segundos podrn informar de las conexiones presentes en distintos tipos
de sinestesia y de si tambin lo estn en la poblacin general. No obstante, tambin es
posible que ambas teoras sean complementarias o que cada una de ellas explique el
mecanismo subyacente a la sinestesia en distintos grupos de sinestsicos.
7 Conclusiones
La evolucin del estudio de la sinestesia en las ltimas dos dcadas ha sido
espectacular y una vez superada la fase de demostracin de la realidad de este
fenmeno, se estn comenzando a abordar interrogantes de gran calado terico y a
desarrollar estrategias adecuadas para su estudio.
El campo de estudio se ha extendido desde la sinestesia grafema-color hasta
otras variantes hasta hace poco inexploradas: sabores evocados por msica (Beeli y
cols., 2005), sabores evocados por las palabras (Ward y Simner, 2003), estructuras
espaciales evocadas por unidades de tiempo (Smilek y cols., en prensa), colores
evocados por la msica (Ward y cols., 2006), etc.
Todo este nuevo conocimiento ayuda a definir globalmente el fenmeno de la
sinestesia al comparar las similitudes de los distintos tipos y as entender si tienen un
origen comn o se deben a procesos diferenciados cuyas caractersticas observables
parecen similares. El creciente aumento del conocimiento a cerca de este fenmeno
tambin est facilitando el estudiar su relacin con procesos de percepcin e
integracin crossmodal y el grado en que percepcin crosmodal en personas no
sinestsicas y sinestesia comparten mecanismos (Sagiv y Ward, en prensa).
50
Iniciativas serias de estudio de los componentes genticos de la sinestesia se
han puesto en marcha y un optimismo generalizado aguarda los resultados de dichas
investigaciones con la esperanza de que proporcionen informacin til para distinguir
entre distintas teoras explicativas y categorizar los distintos tipos de sinestesia y
aquellos otros fenmenos con caractersticas similares pero orgenes distintos.
Por ltimo, cabe mencionar que uno de los temas que menos atencin ha
recibido ha sido el del carcter emocional de la sinestesia. Aunque ha sido
repetidamente descrito en la literatura cientfica, su estudio emprico an no ha sido
abordado. Este es precisamente el objetivo del presente trabajo de investigacin:
iniciar el estudio de las reacciones emocionales asociadas a la percepcin sinestsica
desde un punto de vista emprico para comprobar su existencia, estudiar sus
caractersticas y evaluar si comparte peculiaridades con las reacciones afectivas
experimentadas por personas no sinestsicas.

MOTIVATION OF THIS RESEARCH

Motivation of this research __________________________________________________ 53
The aim of this thesis was to acquire a better understanding of the processes
underlying synesthesia in general and the affective reactions experienced as a side
effect of synesthesia in particular.
A large body of research claims that synesthesia occurs pre-attentively (Smilek,
Dixon, Cudahy & Merikle, 2001; Palmeri et al., 2002; Ramchandran & Hubbard, 2001a)
and a similar amount of research points in the opposite direction (Sagiv & Robertson,
2005; Rich & Mattingley, 2003). We wanted to test these alternative theories and check
whether we could shed some light to this debate. In order to study this issue we
investigated three main questions:
How automatic is synesthesia?
Can synesthesia be suppressed?
Is synesthesia preattentional?

On the other hand, Synesthesia has been characterized as emotional (Cytowic,
1993). Other researchers have also reported that synesthetes inform of a pleasant
feeling when a stimulus they perceive match their internal perception for it and an
unpleasant feeling when it does not match (Ramachandran & Hubbard, 2001b). Thus,
this affective reaction could be though of as a side effect of the synesthesia experience
itself. The second, and more important, aim of our research was to empirically test for
the reality of this affective reaction that has been informally reported, but never
empirically tested, and to study its nature and functioning. Specifically we aimed at
answering the following questions:
Are affective reactions associated to the perception of a synesthetic
incongruence real?
Are they automatic? Can they be ignored? Can they be modified and what
factors influence them?
Are they a ubiquitous phenomenon?
Are they strong enough as to condition co-occurring events?

These two main issues, the role of attention and automaticity on synesthetic
perception and the emotional reactions associated to it, were thoroughly investigated
on two series of experiments, which are reported in the present thesis in two different
sections.
________________________________________ MOTIVATION OF THIS RESEARCH
54
OVERVIEW OF THE RESEARCH
In the First Section we carried out a set of experiments in our synesthete MA to
test the attentional features of her synesthesia. In order to test how automatic her
synesthesia was, in Chapter 1 we carried out a set of Stroop experiments (Stroop,
1935) where we studied the automaticity of her normal perception of colors and how
they influence the naming of her internal photisms and the automaticity of her
synesthetic perception and how it influences the naming of the presented colors. By
comparing the interference produced by each type of perception we tried to deduce
how automatic color perception and photism perception were. These procedures were
extended to another synesthete (PSV) in Chapter 4.
In order to test whether synesthesia could be suppressed we went back to the
experiments carried out in Chapter 1 because they were designed so that we could
also measure Negative Priming. We were then able to test how efficient our synesthete
is to inhibit real colors as compared to photisms. In Chapter 2 we also studied this
issue by carrying out another color naming task where stimuli were hierarchical figures.
By asking her to ignore one dimension and focus on the other we tested whether she
could also ignore the photism elicited by the unattended dimension so that it would not
interfere in her performance.
Last, to test whether synesthesia is pre-attentional we carried out a series of
visual search experiments in Chapter 2. The orientation of a shape made out of letters,
embedded on a full matrix of different letters, has to be searched for discriminated. By
manipulating the perceptual and synesthetic features of the letter making the target
shape we aimed at testing the relative influence of each dimension on the searching
task and the conditions under which the synesthetic features would aid in the search of
the target stimulus.
In the Second Section of this dissertation we turned our interest to the affective
reactions reported to be elicited by synesthesia. Five series of experiments were
carried out.
In order to answer the first question and check whether the usual reports of
synesthetes regarding synesthetically elicited affective reactions informed of a true
phenomenon we carried out a valence rating experiment (Chapter 3) where we
manipulated the color in which a set of words with semantic positive, neutral and
negative valence were presented. By presenting the words in a congruent color (the
Motivation of this research __________________________________________________ 55
same color that a synesthete experiences for a given stimulus) and an incongruent
color (we used the color-wheel opposite to the congruent one) we checked whether the
emotional valence rating was modulated by synesthesic congruence.
We acknowledged that if we found a modulation of synesthesic congruence we
had to check that it was due to the influence of the affective reaction that it produced. In
order to test this and whether the affective reaction was automatic, we conducted a
categorization experiment with a speeded response where again, semantically
emotional words were presented in a congruent and an incongruent color. If an
affective reaction was elicited by the incongruence and it was automatic, we predicted
that it would influence the valence categorization times in a different way for positive
and negative words, leading to an interaction between the two sources of emotional
information: that of the semantic information of the emotion related words, and that of
the affective reaction elicited by the color-photism congruence.
In Chapter 4 we aimed at answering the question about the generalization of
the results we found. We tested three more synesthetes with widely different
backgrounds with the same procedures to check whether the results previously found
were due to our synesthetes idiosyncrasies or to a feature of synesthesia shared by
others. Two of the synesthetes showed similar effects to those shown by MA (although
not as pronounced), whereas another synesthete (PSV) reported strong aversive
reactions but showed no behavioral sign of them. A further two experiments were
carried out to test the possible reasons explaining the lack of correspondence between
PSVs self report of aversive reactions elicited by color-photism incongruence and the
absence of behavioral effects.
In Chapter 5 we tested some factors that could influence the affective reactions
found in previous experiments. We checked whether some colors were more
incongruent than others. Also we checked whether the context in which a color was
presented could influence the perception of that color as incongruent. Last we checked
the effect of the subjective perception of a color when wave length was kept constant
on how incongruent it was perceived. All these variables were tested with a valence
rating design similar to the one used in Chapter 3.
In order to check whether the affective reaction would be of a sufficient strength
as to condition co-occurring events we carried out a classical conditioning experiment
in Chapter 6 with a newly developed implicit learning paradigm (Olson & Fazio, 2001)
in which we checked whether the inconsistency between the photism elicited by the
________________________________________ MOTIVATION OF THIS RESEARCH
56
number signaling the position of a stimulus and the color of the stimulus presented in
that position would elicit a strong affective reaction that would in turn condition attitudes
towards that stimulus.
Last, we checked whether non-synesthetes would also show a modulation of
performance by the implicit inconsistency between their color preferences and the
presented stimuli. Specifically we tested whether non-synesthetes are influenced by
their color preferences when performing tasks for which color is an irrelevant
dimension. Although color preference is somewhat different than preference for a
match between color and photism for a given stimulus, a positive result could lead us to
think that the processes involved in the effect of affective reactions elicited by
synesthetic incongruence could be related to those giving rise to preference effects in
non-synesthetes influencing their performance. Again we used a valence categorization
task where non-synesthetes had to categorize words colored either congruently or
incongruently with their color preferences.
The results of the seven experimental chapters are briefly summarized and
discussed in the General Discussion.

SECTION 1:
SYNESTHESIA AND ATTENTION

CHAPTER 1
AUTOMATIC PERCEPTION AND SYNAESTHESIA:
EVIDENCE FROM COLOUR AND PHOTISM NAMING IN A
STROOP-NEGATIVE PRIMING TASK
1

1
Published in a special issue on synesthesia in Cortex as Lupiez & Callejas (2006), 42, 204-212
.
Chapter 1. Automaticity, Stroop and Negative Priming______________________________
61
ABSTRACT
It is widely assumed that synaesthetic perception is highly automatic, as shown
by Stroop tests. Furthermore, it has been shown that, although automatic, it can be
suppressed leading to Negative Priming. However, these assumptions have not been
consistently investigated, as not many papers have measured Stroop in synaesthesia,
and only one used a Negative Priming procedure. Two experiments were carried out in
a female synaesthete (MA), and 13 control participants, in which numbers and letters
were displayed in colours either congruent or incongruent with MAs photisms. In
contrast to control participants, MA showed significant Stroop effects both when naming
the colours and when naming the photisms (slower RT when naming a colour or
photism that was incongruently coloured versus congruently coloured). For
comparison, we also report a control experiment in which the first letters of colour
names were displayed in either congruent (e.g., B in blue) or incongruent (e.g., B in
red) colours. Significant Stroop and Negative Priming effects were found when a
control group named the displayed colour of these letters. The synaesthesic Stroop
effect shown by MA was greater than that observed in the control experiment when MA
was to name the displayed colour, but smaller when she was to name the photism of
the stimuli. Regarding Negative Priming, MA showed an effect similar to that observed
in the control experiment, but only when she was to name the photisms of the stimuli.
Altogether, these results show that synaesthetic perception is highly automatic and
difficult to inhibit.
INTRODUCTION
Synaesthesia is the experience of an atypical dual perception where a particular
stimulus elicits a normal sensory percept together with another percept which is not the
one usually produced by such stimuli. This additional perception can be categorized as
belonging to the same sensory modality or to a different one (i.e. grapheme-colour
synaesthesia or music-colour synaesthesia respectively). In fact, the most common
type of synaesthesia is the one linking graphemes and colours (Day, 2003) and it is
usually referred to as grapheme-colour synaesthesia. Here the additional perception is
called a photism.
Synaesthesia is not a learned association between colours and specific
graphemes. Recent studies have demonstrated that it is a sensory/perceptual
phenomenon and not a memory based effect (Ramachandran and Hubbard, 2001a).
____________________________________ SECTION 1. SYNESTHSIA AND ATTENTION
62
Synaesthesia is involuntary in the sense that it automatically happens when the
eliciting stimulus is present and it cannot be dismissed at will. Although the eliciting
stimulus is usually physical, it has also been shown that synaesthesia can be caused
by a conceptual stimulus (i.e., by mentally activating a number concept, Dixon, Smilek
et al., 2000).
Although Cytowic (1995) defined synaesthesic percepts as being projected
sensations (e.g., the photism is perceived as an overlay on top of the grapheme in the
case of grapheme-colour synaesthesia), for some synaesthetes these percepts are
experienced in the minds eye (Smilek, Dixon et al., 2002). Synaesthesia is also
durable, meaning that for any particular synaesthete, the synaesthetic associations do
not change over time, they are consistent, and it is also generic as opposed to
elaborated. Specifically, synaesthetes see blobs, lines, spirals and the like but not
elaborated percepts. Synaesthesia is memorable; it is remembered much easier than
the stimulus that elicited it. It is also emotional in the sense of certitude and conviction
that what is perceived is really experienced. Some of these features of synaesthesia
have been used as diagnostic criteria (Cytowic, 1995), in order to differentiate real
synaesthesia from mere crossmodal or unimodal memory associations.
When synaesthesia became of interest to cognitive psychologists, the basic
studies that were carried out tried to show whether it was an automatic phenomenon or
alternatively whether it could be controlled by peoples will. Stroop type studies were
devised to test this automaticity of synaesthetic perception. However, most studies up
to date were carried out with a methodology that did not allow for the analysis of single
responses (i.e. presenting stimuli on cards, and recording RTs with a stopwatch) which
made it difficult to draw some conclusions from them (Mills, Boteler et al., 1999; Mills,
Viguers et al., 2002; Odgaard, Flowers et al., 1999; Wollen and Ruggiero, 1983). More
recent studies, however, have been carried out using more fine-grained methods to
assess this issue (Dixon, Smilek et al., 2002; Elias, Saucier et al., 2003; Mattingley,
Rich et al., 2001) and the common finding points to a large Stroop effect on
synaesthetic perception. Synaesthetes take longer to name the colours of digits or
letters when they are different from the photism perceived for such stimuli. Therefore,
several studies have already shown Stroop effects in synaesthetic perception. A
conclusion drawn from these studies is that photisms are automatically elicited and can
disrupt performance when they are incongruent with the task at hand.
However, it is not clear to what extent photisms activation can be inhibited by a
synaesthete. In this sense, only Odgaard and colleagues (1999) have shown Negative
63
Priming (NP, Tipper, 1985) in synaesthetic perception. The NP paradigm is used to
study the capacity of the attentional system to inhibit information that is not relevant for
the task being performed (Tipper, 2001). This is done by measuring the effect on
subjects performance of previously inhibited information that is now relevant (i.e. a
previous distracter is now used as target stimulus). In their study, Odgaard, Flowers et
al. (1999) asked a synaesthete to name the colour in which a set of numbers were
presented ignoring the photism that such numbers elicited. When the list was ordered
so that the photism of the previous stimulus (which had to be ignored) was now the
colour of the current stimulus, the synaesthete took longer to respond than in those
cases where the target colour had not been previously suppressed. To our knowledge,
this is the only study that showed NP from previously ignored photisms. Furthermore,
in this study a blocked design was used. Therefore, it is still to be shown that ignoring a
photism leads to NP in the following trial, when the effect is measured with the standard
trial by trial procedure.
The above findings suggest that the perception of photisms is highly automatic
although it can be efficiently suppressed when it is detrimental to the task at hand. The
automatic nature of synaesthetic perception seems to be quite a robust finding.
However, the extent to which it can be controlled by the synaesthete is less clear since
it has not been consistently tested. This was one of the aims of the research reported
in this paper.
The general aim of the present study was to explore automaticity of
synaesthetic perception, and the degree of control that synaesthetes are able to exert
over it. In order to do so, two strategies were followed. On the one hand, Stroop and
Negative Priming effects were measured with a trial by trial procedure, in a female
synaesthete, MA, and 13 control participants. One of 12 possible letters or digits was
displayed in each trial, in one of four colours. Participants were required to name the
colour in which the presented letter/digit was displayed, which could be either
congruent or incongruent with MAs photism (allowing us to assess a Stroop effect),
and could be the same or different from the previously ignored photism (allowing us to
assess Negative Priming). On the other hand, MAs performance in the colour-naming
task was compared to her performance in a photism-naming task, in order to compare
the automaticity of response activation from external colours vs. internal photisms.
The effects observed in these two experiments were compared to those
observed in a control experiment where non-synaesthetes were to name the displayed
colour in which the letters presented were the first letters of colour names (henceforth
64
referred to as colour initials). The purpose of this experiment was twofold: on the one
hand, we wanted to reproduce standard Stroop and NP effects in order to ensure the
appropriateness of the procedure being used. On the other hand, the control
experiment had the goal to serve as a baseline condition with stimuli as similar as
possible to those used with MA.
If synaesthetic perception is highly automatic, as previously shown in the
literature, MA should show a Stroop effect in the colour-naming task of our trial by trial
procedure. Furthermore if, despite of being automatic, photism synaesthetic perception
can be easily suppressed when it interferes with the task at hand, MA should similarly
show a Negative Priming effect, as previously reported by Odgaard et al (1999), in a
blocked design. Finally, by comparing MAs photism-naming performance with her
colour-naming performance, a much stronger test of photism automatic perception will
be shown. If, in spite of being automatic, photism perception is less automatic that
colour perception, then a weaker Stroop effect will be observed in the colour-naming
task, where photism interference is measured, than in the photism-naming task, where
colour interference is measured.
EXPERIMENT 1. CONTROL EXPERIMENT
A standard Stroop colour-naming task was adapted for comparison with the
critical experiments in which MAs synaesthesia was investigated. Hence, instead of
presenting the whole colour name, participants were shown the first letters of one of
four colour names displayed in any of the four associated colours, leading to the
Congruent and Incongruent standard conditions of the Stroop procedure.
As mentioned above, the aim of this experiment was twofold. First, we wanted
to make sure that the specific procedure that was used for measuring synaesthetic
Stroop and NP effects in MA was appropriate for measuring the standard Stroop and
NP effects. Second, we intended the experiment to serve as a baseline condition for
comparison with synaesthetic effects.
Furthermore, in order to study the implementation of online control processes to
overcome the interference, apart from NP, the effect of the congruency context was
investigated. Note that this can only be investigated when congruency is manipulated
on a trial by trial basis. Thus, the Stroop effect observed when the previous trial was
congruent was compared to that observed when the previous trial was incongruent. If
65
control processes are dynamically applied depending on online task demands, the
Stroop effect should be bigger after congruent trials, a situation in which the system is
behaving in a less controlled manner (Carter, Macdonald et al., 2000).
METHOD
Participants
Ten control participants voluntarily participated in the experiment for course
credit. They were female psychology students from the University of Granada, and all
of them had normal colour vision.
Stimuli and apparatus
The initials of four colour words were used as stimuli (R, Ro, RO, for rojo red
in Spanish-; V, Ve, VE, for verde green-; Az, AZ, az, for azul blue-; and Am, AM,
am, for amarillo yellow-). Each stimulus was displayed in any of the four different
colours in different trials.
An IBM 14 screen controlled by a 486 computer running MEL software
(Schneider, 1988) was used to present the stimuli. Participants sat at about 50 cm in
front of the computer screen. A microphone connected to a voice-key was used to
record the onsets of the naming responses of the participants. The computer keyboard
was used to record accuracy.
Procedure
Participants were required to name the colour of the stimulus displayed on each
trial. A trial started with a black fixation cross ("+") displayed at the centre of the
computer screen, over a light grey background. After 1000 ms, one of the 12 colour
initials was presented, also at the centre of the screen. The stimuli were
pseudorandomly presented so that neither the same colour nor the same initials were
presented on two consecutive trials.
Participants were instructed to name the colour of the stimulus, while ignoring
the letters. The stimulus was present until participants colour-naming response was
emitted. Then, the experimenter recorded, via a key press, the response given, for
measuring accuracy. Spurious responses due to hesitations or any other misleading
66
activation of the voice-key were also coded, so that they could be discarded from the
analyses.
The experiment consisted of a block of practice trials and 5 blocks of 36
experimental trials each. Participants were prompted to rest between blocks.
Design
For the analysis of the Stroop effect, responses were coded as a function of the
congruency between the colour in which the stimulus was displayed and the initials
being presented. Congruent trials were those in which the colour of the stimulus was
the one designated by the initials being displayed (e.g., RO displayed in red rojo in
Spanish-). Incongruent trials were those in which the stimulus was presented in a
colour different from the one the initials designated (e.g., RO displayed in blue azul
in Spanish-). Additionally, trials were coded according to the congruency of the previous
trial (N-1). Thus, the design had two within-participant factors: N-Congruency (N-
Congruent, N-Incongruent) and N-1 Congruency (N-1 Congruent, N-1Incongruent).
For the analysis of the Negative Priming effect, responses were coded as a
function of the relationship between the colour of the current stimulus (target) and the
initials presented in the preceding trial (distractor). Ignored Repetition trials were those
in which the initials of the preceding trial designated the colour of the current stimulus
(e.g., RO displayed in blue azul in Spanish-, preceded by AZ displayed in yellow
amarillo in Spanish). Control trials were those in which both the target (colour) and the
distractor dimension (initials) were different from the target and distractor of the
previous trial (e.g., RO displayed in blue azul in Spanish-, preceded by VE displayed
in yellow amarillo in Spanish). Thus, the design had only one within-participant
factor: Repetition (Ignored Repetition, Control). Given our restrictions in the colour and
stimuli selection procedure (i.e., no repetition of either the target-colour or the distractor
colour-initials in consecutive trials), the ignored repetition condition was always
incongruent and was also preceded by an incongruent trial (otherwise, either the target
or the distractor would repeat in consecutive trials). Therefore, to have a suitable
condition to compare the results with, only incongruent trials preceded by incongruent
trials were selected for the control condition.
67
RESULTS
The first trial of each block was eliminated from all the analyses. Furthermore,
trials with either incorrect or spurious responses (1.71% and 0%, respectively),
together with correct response trials with RT either faster than 200 ms or slower than
1700 ms (0.06% and 0.86%, respectively) were excluded from the RT analyses. Mean
RTs for each experimental condition were computed, after exclusions, and were
introduced into specific ANOVAs for the appropriate analyses (see Table I). As usual in
the NP literature, the analysis was performed taking participants as the random factor.
Additionally, an item analysis was carried out in order to be able to compare
performance in this experiment with MAs performance in the following experiments.
Table I. Mean RTs (in ms) and error percentages (in parenthesis) for each experimental
condition of the Control Experiment.
N-1 Congruent N-1 Incongruent
N-Congruent N-Incongruent N-Congruent N-Incongruent N-Incongruent
Control Control Control Control Ignored Repetition
608 773 629 742 767
(1.1%) (3.3%) (1.0%) (1.9%) (2.2%)

Analysis of Stroop effect
A 2 (N-Congruency) x 2 (N-1 Congruency) repeated measures ANOVA was
performed on the mean RTs of the control conditions. The analysis showed a highly
significant Stroop effect, as revealed by the main effect of N-Congruency, F
(1, 9)
=61.46,
p<.0001 [F
(1, 11)
=61.38, p<.0001, in the items analysis], showing that RT was 139 ms
faster in the N Congruent condition (the target colour and the initials were congruent)
than in the Incongruent one. It is worth noting that the Stroop effect was modulated by
the congruency of the previous trial, as revealed by the interaction between N-
Congruency and N-1 Congruency, F
(1, 9)
=7.53, p<.05 [F
(1, 11)
=3.92, p=.073, in the
items analysis]. The 165 ms Stroop effect observed when the previous trial was
congruent was reduced to 113 ms when the previous trial was incongruent.
No effect was significant in the analysis of errors. However, as can be observed
in Table I, the pattern of errors mirrored the pattern of the RT data indicating that there
were no speed-accuracy tradeoffs.
68
Analysis of the Negative Priming effect
A repeated measures ANOVA was performed with Repetition (Control, Ignored
Repetition) as the single factor. A standard NP effect was observed: RT was 25 ms
slower in the Ignored Repetition than in the Control condition. The effect was significant
in the participants analysis, F
(1, 9)
=9.94, p<.05, and marginally significant in the items
analysis, F
(1, 11)
=3.81, p=.077. Again, the effect was not significant in the analysis of
errors, although error patterns mirrored the patterns observed for RTs (see Table I).
DISCUSSION
The present experiment fulfilled the main goals for which it was planned.
Significant Stroop and NP effects were obtained in both analyses, taking participants
and items as the random factor. Furthermore, the Stroop effect depended on the
congruency of the previous trials, as has been observed in previous studies (Carter,
Macdonald et al., 2000; Cohen, Dunbar et al., 1990; Kerns, Cohen et al., 2004).
Therefore, the present experiment seems to be appropriate for measuring both Stroop
and NP effects, and the online modulation of Stroop by contextual factors (congruency
of the previous trial) that modulate the relative control with which the task is confronted.
EXPERIMENTS 2a AND 2b. SYNAESTHESIA EXPERIMENTS
In the following section we report data from two experiments in which the
procedure used in the previous experiment was adapted to measure Stroop and NP
priming effects in MA, a grapheme-colour synaesthete, and a group of 13 new non-
synaesthete control participants. In this case, the measured interference will not reflect
the competition between the colour to be named and the initials of such colours
because those stimuli were purposely excluded. Instead, we used letters and numbers
that elicited in MAs mind one of the four colours used in the previous experiment (red,
green, blue or yellow). This way, the Stroop and NP effects would reflect the
competition between the responses elicited by the colour in which the stimuli were
presented and the photism triggered by the graphemes.
MA participated in two different experiments. In a colour-naming experiment she
was to name the colours in which the stimuli were presented, whereas in a photism-
naming experiment she was to name their photisms. Her performance was compared
to that of a group of non-synaesthete participants. The control group only performed
the colour-naming experiment, since they have no photisms to name.
69
As stated above, if synaesthetic perception is highly automatic, MA should show
Stroop effects. Furthermore, if photism perception, although automatic, can be
suppressed when it interferes with the task at hand, MA should similarly show a
Negative Priming effect. Both effects are expected to be absent in the control group, for
which no photisms are perceived.
More importantly, we will be able to compare the relative automaticity of photism
and colour naming activation by comparing MAs performance in the colour-naming and
photism-naming experiment. The stronger and more automatic a dimension is
activated, the more interference will produce in responding to the other dimension.
Therefore, if photism perception is less automatic than colour perception, weaker
Stroop effects will be observed in the colour-naming task than in the photism-naming
task.
METHOD
Participants
A female synaesthete, MA, voluntarily took part in the study. MA is a psychology
student at the University of Granada, who was 21 years old at the time of testing.
Thirteen female psychology students from the University of Granada volunteered to
take part in the same colour-naming experiment as MA, as a control group in exchange
for course credits. All participants had normal colour vision.
Stimuli and apparatus
Instead of presenting the initials of the different colours used as in the previous
control experiment, seven letters and five numbers were presented. None of them were
the first letter of any of the four colours used in the experiment. The selection of stimuli
was made taking into consideration MAs photisms so that three of the stimuli were
synaesthetically perceived as red (E, D and 2), three were perceived as green (J, 3,
13), three as blue (L, S, 6), and the remaining three were perceived as yellow (F, M, 7).
Everything else was the same as in the previous experiment.
Procedure
The colour-naming task that both control participants and MA performed
required them to name on each trial the colour in which the stimulus was displayed,
70
while ignoring the grapheme. Similarly, in the photism-naming task MA was required to
name the photism (i.e., the synaesthetic colour that she perceives), while ignoring its
displayed video colour.
Control participants took part in one single colour-naming session (a block of 36
practice trials and 5 blocks of 36 experimental trials). MA took part in four experimental
sessions: In the first one she performed the colour-naming task. In the second session
she was required to name the photism instead of naming the colour of the stimuli. In
the remaining two sessions MA repeated the same tasks, but in the reversed order
(photism-naming in the third session, and colour-naming in the last one).
Design
The design of both colour-naming and photism-naming tasks was the same as
that of the Control Experiment, both for the Stroop and the NP analysis, although in this
case congruency and repetition effects did not refer to the relation between displayed
colour and initials of colour words, but to the colour-photism relationship. Thus,
Congruent trials were those in which the stimulus being displayed was coloured
according to MAs synaesthetic colours (e.g., D displayed in red or L displayed in blue).
Incongruent trials were those where the video colour of the stimulus was different from
MAs photism (e.g., D displayed either in green, blue or yellow). For the NP analysis, as
in the Control Experiment, Ignored Repetition and Control trials were always
incongruent.
In both the colour naming and photism naming tasks, Control trials were those
in which both the photism and the displayed colour were different from the photism and
displayed colour of the previous trial. Ignored repetition trials were those in which the
target (the displayed colour in the colour-naming experiment, or the photism in the
photism-naming experiment) was the same as the distractor of the previous trials (the
photism in the colour-naming experiment, or the displayed colour in the photism-
naming experiment) (e.g., a D (red photism) displayed in blue, preceded by L (blue
photism) presented in yellow, for the colour-naming experiment, or D (red photism)
displayed in blue, preceded by J (green photism) presented in red, for the photism-
naming experiment). An item analysis was carried out in order to be able compare MAs
performance to that of the control group, where stimuli were taken as the random
factor.
71
In short, the design of the Stroop analysis had two within-item factors: N-
Congruency (N-Congruent, N-Incongruent) and N-1 Congruency (N-1 Congruent, N-
1Incongruent). The design of the Negative Priming analysis had only one within-
participant factor: Repetition (Ignored Repetition, Control).
An additional within-item factor was introduced into the analysis, in order to
compare MAs performance in the colour-naming and photism-naming tasks to that of
the control group in the colour-naming task. This factor is called Experiment: Control-
Group (Colour-Naming), MA (Colour-Naming) and MA (Photism-Naming.
RESULTS
As in the Control Experiment, the first trial of each block was eliminated from all
the analyses. Furthermore, trials with either incorrect or spurious responses (0.44%
and 0% for the control group, 0.88% and 3.43% for MA colour-naming, and 1.20% and
0.70% for MA photism-naming, respectively), together with correct response trials with
RTs faster than 200 ms or slower than 1700 ms (0.04% and 0.13 % for the control
group, 0% and 0.04% for MA colour-naming, and 0% and 0% for MA photism-naming,
respectively) were excluded from the RT analyses. Mean RTs for each item and
experimental condition were computed, after exclusions, and were introduced into
specific ANOVAs for the appropriate analyses (see Table II).
Table II. Mean RTs (in ms) and error percentages (in parenthesis) for each experimental
condition for the Control Group Colour-Naming Experiment, and for MA Colour-Naming
and Photism-Naming Experiments.

N-1 Congruent

N-1 Incongruent

N-Congruent N-Incongruent

N-Congruent N-Incongruent N-Incongruent

Control Control

Control Control Ignored Repetition
Control Group 656 660 652 654 666
Colour-Naming (0.3%) (1.1%) (0.5%) (0.2%) (0.3%)
MA 749 1108 786 1037 1012
Colour-Naming (0.0%) (0.0%) (0.0%) (0.0%) (3.1%)
MA 784 848 786 797 852
Photism-Naming (0.0%) (2.1%) (1.4%) (1.1%) (0.0%)
72

Analysis of Stroop effect
A 3 (Experiment) x 2 (N-Congruency) x 2 (N-1 Congruency) repeated measures
ANOVA was performed on the mean RTs of the control conditions (i.e., excluding the
ignored repetition condition). Apart from the main effect of Experiment, F
(2, 22)
=82.13,
p<.0001, the analysis showed a highly significant Stroop effect, as revealed by the
main effect of N-Congruency, F
(1, 11)
= 35.29, p< .0001. However, as expected, the
Experiment x N-Congruency interaction was also significant, F
(2, 22)
=84.23, p<.0001,
showing that the Stroop effect was only present in MAs performance (F<1 for the
control group). As shown in Graph 1, the Stroop effect that was observed in MAs
performance followed the same pattern of that observed in the Control Experiment,
with a larger Stroop effect for the congruent context (N-1 congruent) (in fact, the N-
Congruency x N-1 Congruency interaction was also significant, F
(1, 11)
=5.10, p<.05).
This contrasts with the Control participants data where no Stroop effect was observed.
550
650
750
850
950
1050
1150
N-1
Congr u
N-1
Incongr u
N-1
Congr u
N-1
Incongr u
N-1
Congr u
N-1
Incongr u
N-1
Congr u
N-1
Incongr u
Cont r ol Gr oup:
Colour -Naming
MA: Colour -
Naming
MA:
Phot is m-Naming
Cont r ol
Exper iment
Synaes t hes ia Exper iment s (Exper iment 2a & 2b) (Exper iment 1)
N
a
m
i
n
g

T
i
m
e
s

(
m
s
)
Congruent Incongruent

Graph 1: Stroop effects in MA's performance, in the colour-naming and photism-naming
experiments, as compared to the control group of non-synaesthetic participants. For
comparison, data from the Control Experiment are also provided.
In order to have a closer look to the pattern of Stroop effects observed in MAs
performance in comparison to the standard Stroop effects observed in the Control
Experiment, a new 3 (Experiment) x 2 (N-Congruency) x 2 (N-1 Congruency) ANOVA
was performed. In this case, the three levels of the variable experiment were: Control
Experiment, MA/Colour-Naming, and MA/Photism-Naming. The analysis showed that a
significant Stroop effect, F
(1, 11)
=70.76, p<.0001, was different across Experiments, F
(2,
73
22)
= 55.73, p< .0001. Planned comparisons showed that, compared to the control
experiment, the Stroop effect was greater in MAs colour-naming experiment, and
smaller in MAs photism-naming experiment (both ps<.001). Furthermore, the Stroop
effect was again significantly modulated by the congruency of the previous trial, F
(1, 11)
=
7.86, p< .02. It is important to note that this modulation was similar in the three
experiments, as shown by the non significant three-way interaction (F<1) (see Graph
1).
No analysis was performed in the error rates, given that very few errors were
made (see Table II).
Analysis of the Negative Priming effect
A 3 (Experiment) x 2 (Repetition) repeated measures ANOVA was performed on
the mean RTs. Apart from the main effect of Experiment, which was again significant,
F
(2, 22)
= 89.02, p< .0001, the interaction between Experiment and Repetition was
marginally significant, F
(2, 22)
=3.19, p=.061. Planned comparisons showed that the only
significant NP effect was that found in MAs photism-naming experiment, F
(1, 11)
=11.29,
p<.01 (p>.25 in any of the other two cases).
550
650
750
850
950
1050
1150
Control
Group:
Colour-
Naming
MA:
Colour-
Naming
MA:
Photism-
Naming

Control
Experiment
Synaesthesia Experiments (Experiment 2a &
2b)
(Experiment
1)
N
a
m
i
n
g

T
i
m
e
s

(
m
s
)
Control Ignored Repetition

Graph 2. Negative Priming effect in MA's performance, in the colour-naming and
photism-naming experiments, as compared to the control group of non-synaesthetic
participants. For comparison, data from the Control Experiment are also provided.
As we did with the Stroop analysis, a new 3 (Experiment) x 2 (Repetition)
ANOVA was performed, now taking Control Experiment, MA/Colour-Naming, and
MA/Photism-Naming as the three levels of the variable Experiment. The NP effect was
74
significantly different across experiments, F
(2, 22)
=3.96, p<.05. As compared to the
Control Experiment, MAs RTs showed a smaller NP effect in colour-naming, F
(1, 11)
=
5.81, p<.05, whereas she showed a comparable NP effect in the photism-naming
Experiment, F
(1, 11)
=2.44, p>.10. In fact, as can be observed in Graph 2, the NP effect
shown by MA in the photism-naming Experiment was numerically larger than that
observed in the Control Experiment; in contrast, MA showed no NP effect in the colour-
naming Experiment (in fact she showed a non-significant effect in the opposite
direction). Importantly, these effects shown by MA for the two tasks were significantly
different from each other, F
(1, 11)
=4.90, p<.05.
DISCUSSION
As one might expect, participants in the control group did not show any
significant difference in their colour-naming performance depending on whether the
letters and numbers were displayed in the same colour as MAs photism or in any other
colour. In contrast, MA took much longer to name the same colours when the stimuli
elicited a photism different from the colour in which it was presented, than in a
congruent condition in which photism and colour were the same. This synaesthetic
Stroop effect was observed both when MA was required to name the (external) colour
and when she was asked to name the (internal) photism elicited by the stimuli.
However, the effect was much larger in the former case as MAs responses were
especially slow when she had to name the colour of photism-incongruent letters/digits,
but not when naming the photism of colour-incongruent letters/digits.
As reasoned in the introduction, these results are the opposite to what one
would expect if the activation of the response appropriate to the photism were less
automatic than that of the colour being displayed. Consequently, the pattern of data
suggests that the activation of a colour-naming response might be more automatic
when it is produced by the identification of a letter or number than when it is produced
by the processing of a wavelength of light. Therefore, these results add to the literature
showing high automaticity in synaesthetic colour perception.
The priming results are in line with this interpretation. MA did not show any sign
of Negative Priming in the colour-naming experiment. That is, contrary to what could be
expected, she was not slower at naming the colour of an incongruently coloured
stimulus when this colour was the same as the photism of the previous stimulus,
compared to a control situation in which the previous photism was different from the
current to-be-named colour. In fact, she showed some facilitation, although not
75
significant. These results could reflect the difficulties that MA had at inhibiting the
response associated to the photism, when required to name the colour (colour-naming
task). The fact that MA was slower in naming the colour of incongruently coloured
stimuli than in naming the photism of the very same stimuli (see Graph 1) is in
agreement with that conclusion. Therefore, MA seems to have special difficulties for
suppressing naming responses elicited by her synaesthetic perception of letters and
numbers, whereas she seems to have not much difficulty in suppressing colour naming
responses when required to name the photism.
At first, these results are at odds with those of Odgaard et al. (1999), since their
synaesthete L showed NP in an experiment where she had to name the displayed
colour of a set of numbers ordered in a way that the colour of each one was the same
as the photism of the previous item. This effect was equivalent to the Stroop effect
found for colour names. As already mentioned, we did not find any hint of a NP effect in
the colour naming experiment performed by MA. One possible explanation could be
that MAs grapheme-colour synaesthesia operates differently than does Ls. Whereas
MAs synaesthesia seems to be of a high level (Ramachandran and Hubbard, 2001b),
we do not have data to conclude whether Ls synaesthesia is of low or high level.
Alternatively, it might be argued that the absence of NP in the colour naming
performance of MA, but not in her photism naming performance, could be due to MAs
markedly long RTs in the former condition. However, this is unlikely because in fact NP
is usually greater (nor smaller) when accuracy is emphasized, in spite of having longer
RTs (Li, 1996; Neill and Westberry, 1987).
Several other differences could explain such inconsistencies in the results. In
the first place, the methodologies used in both experiments are considerably different.
Whereas we used a trial by trial procedure in which response latencies were measured
for each stimulus, Odgaard et al. (1999) measured the time taken to complete a 13-
item list. An important consequence of the current procedure was that we were able to
manipulate the different conditions within-blocks while they had to manipulate them
between-blocks. The blocked manipulation of the two conditions could be boosting the
adoption of a more controlled task set that would facilitate the observation of NP
effects, even in the colour naming task. In fact, Odgaard (personal communication)
used the blocked design in order to maximize the conditions under which a NP effect
was likely to be obtained.
A different explanation for the inconsistencies in the results has to do with the
emotional state produced by the task. Synaesthetes usually report being disturbed by
76
incongruently coloured stimuli. This disturbance is easily appreciated in MA when she
is performing a common Stroop-type task and has been shown to influence her
performance on behavioral tasks (Callejas, Acosta & Lupiez, submitted, Chapter 3).
Such a state of anxiety induced by the task could interfere with inhibitory processes in
MAs performance. In fact, MA reported that she did not like doing these kinds of tasks,
because they made her anxious. Fox (1994) has reported that Negative Priming can
disappear in trait-anxious individuals when they have to ignore threat-related
information, as well as non threat-related distracting information. Thus, the anxiety
induced by the colour naming task could have caused the lack of NP on MAs
performance on this task (interestingly, MA was not distressed by the photism naming
task). However, in order to take this hypothesis as a possible explanation for the
differences, several things had to be tested. The study carried out by Fox (1994) refers
to trait-anxious individuals as opposed to state-anxious individuals. It is not clear
whether or not the same results would be found with this state measure. Another
problem for this interpretation is that we do not know the emotional reactions
experienced by L when she underwent the experiments reported by Odgaard et al.
(1999).
Alternatively, it could be argued that MA might have a defective control system
that would cause a lack of inhibition of irrelevant information, and therefore an absence
of NP. Two facts disprove this hypothesis. An incapacity to inhibit irrelevant information
would lead to a substantial increase in the percentage of errors for the control condition
of the NP analysis. However, MA performed those trials without making one single
mistake. More importantly, NP was found on MAs performance on the photism-naming
task, thus proving that MA does not have inhibition deficits.
In fact, the NP effect found in the photism-naming experiment is even larger
than that shown by control participants in the control experiment. This finding is
especially interesting when compared to the effect obtained on the colour-naming task,
where a non significant trend towards facilitation was observed instead. This
dissociation could be explained on the basis of the relative strength of the activation of
the colour responses associated to the wave length vs. the photism. In short, the
pattern of results could be explained by arguing that the colour response associated to
the photism is either more strongly activated than that associated to the wave length of
the stimulus, or has a privileged access to consciousness.
A much stronger activation of the response associated to the photism would
account for the pattern of results found in the two experiments. The activation of the
77
response associated to the wave length colour is not so automatic (or has a less
privileged access to consciousness). Therefore, this distracting dimension can be
easily inhibited on trial N-1 (when the task is to name the photism), thus not interfering
with the photism naming response. In fact, its interfering effect was practically
eliminated when enough control was exerted, as on trials preceded by incongruent
trials (see Graph 1). Consequently, this effective inhibition led to the NP that was found
on trial N in these conditions (see Graph 2).
On the contrary, given a strong activation of the photism response (or a
privileged access to consciousness), a great deal of inhibition would be necessary to
prevent its disrupting influence on the response on trial N-1, when the task is to name
the wave length colour. The fact that the Stroop effect observed in these conditions was
reduced from 359 to 251 ms when the preceding trial was incongruent might be
interpreted as evidence that the control system was acting by inhibiting the disrupting
influence of the photism colour response (the distractor in this task). However, a strong
Stroop effect (greater than that observed in the control experiment) was still observed
in this effective-control situation. This ought to be interpreted as evidence that, in spite
of the inhibitory control being applied, the photism colour response could not be
effectively inhibited. Consequently, no NP was found from the photism colour of the
preceding trial on the wave length colour response to the current trial.
Taken together, the Stroop and NP results consistently point to a high degree of
automaticity and strength of synaesthetic perception. The reaction time patterns could
suggest that such perception is, either stronger and more automatic than the one
related to external visual stimuli or has a privileged access to consciousness

CHAPTER 2
THE ROLE OF ATTENTION IN SYNESTHESIA
Chapter 2. The Role of Attention in Synesthesia __________________________________
81
ABSTRACT
A currently hot topic in the grapheme-color synesthesia research is the role of
attention in the binding of grapheme and color. While some researchers suggest that
conscious perception of the inducing stimulus is necessary for the elicited photism to
be experienced, others posit that the photism can be experienced even before the
inducer is consciously recognized. We present the results from five experiments aimed
at testing whether synesthetic photisms aid visual search in our synesthete MA, and
whether focusing attention on one dimension of the inducer stimulus can prevent
incongruent information from the ignored dimension from affecting performance.
Results show that synesthesia does not aid in the search for the target stimulus
although it influences performance by varying the relative similarity between target and
distracters and among distracters. We also find that ignoring the dimension of a
stimulus that elicits an incongruent photism is not sufficient to prevent that
incongruence from influencing performance.
INTRODUCTION
An important issue on synesthesia research is the role of attention on
synesthetic perception. That is, whether synesthetic percepts are automatically
generated or rather attention is necessary for synesthetic percepts to be activated.
Furthermore, in any case, a related issue is the degree in which synesthetic perception
is subject to the control and intentionality of the synestehete. Several papers have now
been published where different positions in the debate about the necessity of attention
for synesthesia are defended. While some defend the perceptual reality of photisms
(Smilek, Dixon, Cudahy & Merikle, 2001; Palmeri et al., 2002; Ramachandran &
Hubbard, 2001a) other groups defend that attention is necessary for synesthetic colors
to appear (Mattingley, Rich, Yelland & Bradshaw, 2001; Sagiv & Robertson, 2005). As
surprising as it might seem, five years have passed since the first papers proposing
one view or the other were first published and a consensus is yet to come. Each
laboratory validating their own results and discrediting those pointing in a different
direction on the basis of methodological problems, statistical problems, nature of the
synesthetic population or rationale used to test individuals or groups, all that has been
put forward up to now is more evidence for each point of view from the corresponding
lab. Far from aiming at an exhaustive revision of the published data, a brief summary
will be given of the data favoring each view.
___________________________________ SECTION 1. SYNESTHESIA AND ATTENTION
82
Binding of grapheme and color is pre-attentive
This idea was first put forward by Ramachandran and Hubbard (2001a) and
Smilek et al. (2001). The first group carried out a task where participants had to
segregate a shape from a background of distracters. Importantly, both distracters and
items making up the shape were letters that induced different color photisms. Using a
blocked design in which participants knew what letter would be making up the
embedded shape, they found that when the display was presented for one second,
synesthetes correctly categorized the form of the shape in 81% of the trials while
controls did so only in 59% of the trials. While controls performed better than chance
(there were four possible forms), synesthetes were significantly better. Ramachandran
and Hubbard (2001a) concluded that this proved the genuinely perceptual nature of the
phenomenon. To add evidence in favor of this idea they carried out another task in
which they presented a grapheme at an eccentricity of about 7 from fixation
(Ramachandran & Hubbard, 2001b). When the digit was flanked by four other digits
(i.e. on both sides, on top and below it) controls participants found it difficult to identify
the central digit. Synesthetes on the other hand were significantly better at doing so.
They claimed that the colors elicited by target and distracters aided in the identification
of the otherwise unrecognizable target.
On the other hand, Smilek et al. (2001) carried out a visual search task where a
synesthete had to find a dark grey target digit surrounded by distracter digits. Critically,
the background color in which the items were presented was colored congruently or
incongruently with the photism elicited by the target digit. They found that the
synesthete took longer to find the target digit when it was presented against a
congruent background color. This finding was interpreted as showing that digits were
experienced with their elicited color and therefore lost in the similar color background.
They also performed an identification task in which masked digits were again
presented in a congruent or incongruent background color. Again identification of
masked digit was better when presented against an incongruently colored background.
Although statistically better in the incongruent color background, it is worth mentioning
that in the worst condition their synesthete still correctly identified 88% of the letters
presented for 32ms and masked for 1000ms.
Subsequent experiments have added data favoring this position. Palmeri, et al.
(2002) asked a synesthete to look for a white target digit presented amongst a variable
size set of distracters. When both target and distracter elicited different synesthetic
83
colors they found that the slope of the synesthetes search function was significantly
flatter than that of the control participants. However, when the distracters were
characters not eliciting any color, his slope resembled that of the control participants.
Authors concluded that, although perceptual in nature, synesthetic colors were not like
true colors in that they did not produce pop-out. Instead they proposed that the search
was a serial-like one in which synesthetes found it easier to reject distracters on the
basis of their color and thus would not search back in those areas. We will come back
to this point in the next section. Recently Laeng, Svartdal and Oelmann, (2004) carried
out a study similar to that of Palmeri et al. (2002) in which both target stimulus and
distracter elicited a synesthetic color. He replicated previous results as searching was
much faster when the eliciting color of the target was different from the eliciting color of
the distracters (being both of them presented in achromatic shade). He then went
ahead and reanalyzed the data as a function of the eccentricity from fixation with which
the target stimulus had appeared in the searching display. Surprisingly, faster reaction
times were due to those trials where the target was close to fixation. When it was
presented in an area 3 to 6 from fixation, the synesthete was faster detecting it than
the control participants. In contrast, for eccentricities larger than this, the responses of
the synesthete did not differ from those of non-synesthetes. They therefore concluded
that since attention was allocated at fixation point prior to the appearance of the
searching display and reaction times differed only for those targets appearing close to
fixation, the differences found between their synesthete and controls were not due to
the pre-attentive pop-out of the photism but to the allocation of attention to the area
where the target was presented.
Similar results have been observed in out lab (Lupiez & Callejas, unpublished
data). We presented either two or four characters at four possible locations (to the left,
right, below or above the fixation point). One of the characters was always a number,
whereas the remaining characters were letters. One synaesthete (MA) was asked to
search for the number target and press the corresponding key depending on whether it
was odd or even. In the condition where the target elicited a photism different from that
elicited by the distracters the slope of the search function was reduced to 20 ms per
item, as compared to the 52 ms obtained when both the target and the distracters
elicited the same photism. However, in a subsequent experiment, in which a new
condition was added, where the popping out element was one of the distracters, the
effect disappeared.
84
Binding of grapheme and color requires attention
Opposing the already mentioned authors there are two groups proposing that
attention is necessary for the binding of (synesthetic) color to grapheme. Mattingley et
al. (2001) reported a study where, by means of masking a prime stimulus to render it
unconscious (i.e. a grapheme), they showed no influence of such stimulus on the
subsequent naming of a color patch in a group of 15 synesthetes. Importantly, when
the prime was consciously perceived (500ms display time) they found a congruency
effect between the photism associated with the prime stimulus and the color of the
patch (i.e. slower color naming times when the photism elicited by the prime did not
match the color of the patch to be named). Based on these results, they concluded that
attention and conscious perception of stimuli is necessary for the synesthetic color to
appear.
Adding to this they performed another series of experiments (Rich & Mattingley,
2003) where they presented the group of synesthetes with complex displays with a
Navon-like structure (i.e. global forms made out of local forms). Global and local forms
could be made out of the same stimulus (thus inducing the same photism) or different
stimuli (thus inducing two different photisms). In a first experiment they asked
participants to name the color in which the display was presented and found that when
the color was incongruent with both eliciting stimuli (i.e. global stimulus and local
stimulus) participants were much slower than when the color was congruent with both
eliciting stimuli (i.e. both global and local stimuli). Those conditions where one
dimension was congruent and the other was not (i.e. either the global form was
congruent with the displayed color and the local form incongruent or vice versa) yielded
intermediate naming times. In order to elucidate whether attention could modulate this
pattern of results, they carried out a second experiment in which they instructed
participants to focus attention either on the global level or the local level. Now only
both-levels congruent or attended level congruent-ignored level incongruent conditions
were presented. Under these circumstances they found that the magnitude of the
congruency effect (i.e. difference between both-levels congruent and attended level
congruent-ignored level incongruent) showed the following pattern. When participants
were instructed to attend to the global shape, the congruency effect did not change
across experiments. When participants had to attend to the local shapes, although the
congruency effect was still statistically significant, it was reliably smaller than the effect
produced by the same conditions in experiment one. Authors concluded that, although
actively ignored letters can induce a synesthetic color that interferes with naming of the
85
displayed color [], actively ignoring a letter can reduce its synesthetic effect
(pp.1797).
Sagiv and Robertson (2005) carried out a set of experiments on a pair of
synesthetes to test the pre-attentive view point. They designed a visual search task in
which rotated letters appeared as non-sense symbols therefore not eliciting any
photism. In a different experiment they rotated back the target stimulus so that now it
was an L shape that did elicit a color. They tested each synesthetes performance
when neither distracter nor target gave rise to a synesthetic color and when target but
not distracter did elicit a color. Results showed that the searching functions for target
present trials were not different when the target induced a synesthetic color than when
it did not. This is the same result found by Palmeri et al. (2002) in their second
experiment and, although they did not take it to mean that photisms are not perceptual
in nature, Sagiv and Robertson concluded that synesthetic colors do not pop out but
behave in a way that better resembles a type of guided search (Sagiv, Heer and
Robertson, 2006). Interestingly, these two synesthetes were also tested by Hubbard
(Hubbard, Arman, Ramachandran & Boynton, 2005) and they found that their
performance on the embedded figures test was better than that of control participants.
They also run the crowding task and now only one of the synesthetes was better than
the matched control group. As Hubbard et al. (2005) suggest, this might be showing
that synesthesia is a heterogeneous quality.
Even more interesting is to find that, although Sagiv and Robertson (2005) did
not find an interaction in the reaction times for detection of target present trials as a
function of set size and whether the target was or not a synesthetic inducer, they did
find a main effect of target type in one of the two synesthetes. Overall, this synesthete
showed faster response times when the target stimulus was an inducer. This
synesthete was the one performing better than the matched control group in the
crowding task. So far, this is the only time the same synesthete is tested with different
paradigms by different researchers and results seem to be adjustable. Whereas
Ramachandran and Hubbards embedded figures task might be too lax of an evidence
to prove perceptual effects, Sagiv and Robertsons task might be overlooking some
interesting patterns of results.
Several authors have criticized Mattingley and Richs work on the ground of
methodological problems. It is widely accepted that synesthetes are not equal (Dixon,
Smilek & Merikle, 2004; Dixon & Smilek, 2005; Hubbard et al., 2005; Ramachandran &
Hubbard, 2001b; Smilek, Callejas, Dixon & Merikle, in press) and that different kinds of
86
synesthesia might give rise at different levels of the visual processing. It is also clear
that different synesthetes show characteristic patterns of data while performing the
exact same task. Mattingley and Rich refuse to take this fact into account and perform
all their studies averaging group data from a relatively large sample of synesthetes with
the consequent contamination by different patterns of performance and difficulty in the
replicability of their findings. Blake, Palmeri, Marois, and Kim (2005) and Smilek, Dixon
and Merikle (2005a) also mention some specific flaws of the priming study that have to
do with the start up assumption that color can be perceived without awareness and
that, if so, it can prime a subsequent response or the possible lack of experimental
power to detect a difference that is nominally equal in the unconscious priming
condition than the statistically significant difference in the control priming experiment.
We also have some comments about their Navon-like stimuli experiment that will be
discussed in more detail in a later section.
More evidence has been added to both points of view in the last years (Edquist,
Rich, Brinkman & Mattingley, 2006; Kim, Blake & Palmeri, 2006; Mattingley, Rich,
Yelland & Bradshaw, 2006; Rich & Mattingley, 2005; Wagar, Dixon, Smilek & Cudahy,
2002) but there is still no glimpse of a consensus. Until methodologically spotless
experiments are designed and careful analysis of the individual patterns of results are
carried out, a solution to this controversy will not arrive. A good attempt to do so is the
study by Sagiv and Robertson (2005) where the same subjects that have been
extensively studied by Hubbard et al (2005) have been tested on a visual search task
similar to the one used by Palmeri et al (2002) or Laeng et al (2004) and have yielded
somewhat conflicting results.
We carried out a set of experiments to study the role of attention in synesthetic
perception, by testing how our synesthete MA behaved in situations similar to those
already described. We designed a set of visual search tasks similar to those used by
Ramachandran and Hubbard (2001a) where we compared the potential benefit of
searching for a different shape with the benefit of searching for a different (synesthetic)
color. Also we designed another task similar to that utilized by Rich and Mattingley
(2003) with some methodological improvements.
EXPERIMENT 1. VISUAL SEARCH WITHOUT SEARCHING CUES
The goal of this first experiment was to explore extra features of MAs
perception. In particular, we were interested in learning more about the benefits that
87
perception of the photisms might have in MAs performance when perceiving a difficult
stimulus. In order to do so we designed a visual search task in which we combined
perceptual similarity with synesthetic similarity to study the relative contribution of each
dimension on searching for a camouflaged target.
Taking advantage of the fact that there were several graphemes eliciting the
same photism in our synesthete MA, we chose 2 of them that were substantially
different in their perceptual shape but elicited the same color (i.e. both M and N
elicited yellow). Also, we chose another grapheme that was perceptually equal (i.e. W
as a mirror image of M) but elicited a substantially different color (purple). Instead of
asking participants to search for a target stimulus, we constructed a rectangular shape
out of it and asked participants to discriminate its orientation (i.e. vertical vs. horizontal).
The embedded figures made out of these three letters were quasi randomly distributed
across the display and the other two letters were presented as distracters.
A previous pilot experiment had shown that the searching times for an
embedded shape conformed by Ns were considerably faster than those for the
shapes made out of Ms or Ws in a group of non-synesthetic controls. We therefore
predicted that control participants would find a benefit when the shape was made out of
the N letters because of its perceptual features. However, we predicted that if MA
could take advantage of her synesthetic colors, she would be faster finding a shape
made out of Ws since, for her, it was clearly different from the other letters according
to its elicited color. We also expected the shape made out of Ms to produce the
slowest reaction times since it had the same perceptual shape than one distracter and
the same elicited color than the other distracter. Last, we expected N to fall in
between, closer to M or to W depending on which dimension of the stimulus would
affect MAs performance in a stronger manner (i.e. unique shape vs. unique synesthetic
color).
METHOD
Participants
MA and a group of 10 undergraduate psychology students participated in this
experiment. They all had normal or corrected to normal vision and the mean age of the
group was 21. All of them were women (including MA).
88
Stimuli and material
Three letters were chosen as stimuli for this task based on their perceptual
features and MAs photism for them. As can be seen in Figure 1, the letters we used
were M, N and W. The first two letters elicited a yellow photism while the last one
elicited a purple one. Therefore, photism-wise, M and N were very similar and W
was very different. On the other hand, shape-wise, M and W were the exact same
shape after it had been horizontally flipped and N was fairly different.
The whole display had a total of 81 letters (9 rows and 9 columns) and at a
viewing distance of 53cm subtended 10.78 horizontal and 9.71 degrees vertical. Each
individual letter was 0.65-0.86 horizontal and 0.86 vertical. The rectangle was formed
by 5 letters on the long side and 3 letters on the short side. Therefore it was about
5.94 x 2.92. The rectangles inside was filled with randomly chosen letters other than
the one making it. For each display, the letter used to build up the rectangular shape
was never also used as distracter (i.e. a display that had a rectangle made out of Ms
did not have that letter anywhere else other than in the outline of the rectangle).

vertical rectangle made out of Ns ; in the middle one the horizontal rectangle is made
out of Ws and in the right one the horizontal rectangle is made out of Ms .
Stimulus presentation as well as data collection was carried out on a Pentium
computer running E-prime software (Schneider, Eschman & Zuccolotto, 2002).
Participants responded by pressing one of two letters in a computer keyboard. A 14
computer screen was used and auditory feedback was given by means of headphones.
Design and Procedure
The task that participants were to perform consisted on a visual search in which
the searched target was a rectangular shape made out of 12 letters embedded in a 81-
89
letter display (9 rows and 9 columns). The rectangle was made out of one of the three
letter types used in the display and distracters were the other two letters, randomly
distributed across the display. The rectangle could be either horizontally or vertically
positioned and could appear in four different locations to prevent participants from
expecting where to find it. Participants task was to find the rectangle and press either
the z or the b key depending on its orientation. MA).
In each trial a fixation point was presented for 800ms and followed by the
search display. After a response was given, feedback was emitted if a mistake had
been made. Otherwise, the next trial began.
Four blocks of trials were presented with 24 trials each (one per location,
position and letter condition). Participants were instructed to respond as fast as
possible whether the embedded rectangle was horizontal or vertical and to do so
without making mistakes.
RESULTS AND DISCUSSION
Reaction time analysis
Incorrect responses (7.2%) were excluded from the reaction time analysis.
Reaction times were considerably slow for this task and also relatively variable among
participants. We therefore used individual standard deviations as criteria to identify and
exclude outliers. All response times slower or faster than 2 SD of each participants
mean were excluded from further analysis. Blocks were used as random factor. Control
groups scores were averaged across participants for each experimental block and
subsequently compared to MAs scores. Mean reaction times and error rates per group
and condition can be seen in Table I.
Table I. Experiment 1. RT (and percentage correct) per group and experimental condition.
Letter M N W
MA 4451.92 (11.1%) 2750.03 (3.1%) 4687.28 (3.3%)
CTRLS 3750.43 (8.8%) 3208.91 (4.9%) 4001.65 (4.1%)

A 2 (Group) x 3 (Letter) ANOVA was performed on the averaged data and
individual ANOVAs were carried out when necessary. The general analysis informed
that there were no differences between groups (F
(1, 3)
=2.85, p=.189). Letter N was
90
easier to find than any other letter (F
(2, 6)
=10.64, p=.011) and this was so for control
participants (F
(2, 6)
=17.60, p=.003) as well as for MA (F
(2, 6)
=7.22, p=.025). Planned
comparisons showed that for MA, searching for a purple rectangle among yellow
distracters (i.e. W condition) was no different than looking for a yellow rectangle
among yellow and purple distracters (i.e. M condition) (F<1). However, looking for a
rectangle where the letter forming it was perceptually different than the distracters (i.e.
N condition) was in fact faster than the other two conditions (F
(1, 3)
=7.017, p=.077 and
F
(1, 3)
=14.69, p=.031 respectively). Controls did not show any difference between M
and W. They were both more difficult to find than N (both ps<.02).
1500
2500
3500
4500
5500
M N W
S
e
a
r
c
h
i
n
g

R
T
s

(
m
s
)
Controls MA

Graph 1. Experiment 1. Reaction times for MA and the Control group as a function of the
letter that formed the target stimulus. Error bars denote standard errors of the mean.
As can be observed in Graph 1, this pattern of results was the same as that
found with the control participants. It seems then that MA was not benefiting from the
extra dimension of the stimuli (i.e. color) that only she perceived. Her pattern of results
showed that she obtained a benefit due to the letter shape, being it of the same type as
the one obtained by the control participants. Nevertheless, this particular shape should
be more difficult to find for her since it elicited the same color as one of the distracter
letters.
Accuracy analysis
The same 2 (Group) x 3 (Letter) ANOVA was carried out for the percentage
correct data. None of the main effects or the interaction reached significance. As can
be seen in Graph 2, since no time pressure was given to participants, the overall error
rates were fairly low for all conditions for MA as well as for the control group.
91
0%
5%
10%
15%
20%
M N W
E
r
r
o
r

R
a
t
e
s

(
%
)
Controls MA

Graph 2. Experiment 1. Percentages of correct responses for MA and the Control group
as a function of the letter that formed the target stimulus. Error bars denote standard
errors of the mean.
From the reaction time data we concluded that colors (photisms) do not appear
automatically to MA as she fixates in a display, and therefore she does not benefit from
the extra information that they could provide. However, the fact that no influence of
letter was found in this experiment does not rule out the possibility that she is in fact
able to use the color information for this type of task as long as she knows what she is
looking for. That is, if she knows what letter is forming the display, as was the case in
Ramachandran and Hubbards experiment (2001a), she might tune her search to only
look for the stimuli eliciting that color. If this was so, we would conclude that colors do
not pop-out of the display but they could be able to guide search.
In order to test this hypothesis, we conducted Experiment 2 in which, before the
search display was presented, a cue was provided informing about the identity (and
therefore color for MA) of the stimuli forming the rectangle for that particular trial.
EXPERIMENT 2. VISUAL SEARCH WITH SEARCHING CUES
In this new experiment, we added a visual cue that indexed the identity of the
letter that had to be searched for. We expected that giving this piece of information to
participants in general would enhance their performance since their search had a
predefined objective. Also, we expected a differential benefit for MA since now she had
an additional piece of information that should aid her search. By knowing the identity of
the target stimulus she would also know the color she had to look for.
92
METHOD
Participants
MA and the same group of 10 students who participated in the previous
experiment performed this experiment.
The task to be performed was the exact same one as in the previous
experiment with the only difference of a cueing letter appearing right before the search
display. The fixation point used in Experiment one was replaced in this experiment by
the letter, which was also presented on the screen for 800ms. This letter informed of
the identity of the stimulus that would be making up the rectangular shape. This time,
six blocks of trials were presented.
We again used a 2SD cut off to eliminate outliers for each participant. Incorrect
responses (6.4%) were also excluded. Again, blocks instead of participants were used
as random factor. Mean reaction times and error rates per condition can be seen in
Table II.
Table II. Experiment 2. RT (and error rates) per group and experimental condition.
Letter M N W
MA 1897.65 (29.1%) 1284.01 (8.3%) 1762.25 (7.5%)
CTRLS 1987.98 (13.2%) 1389.96 (9.2%) 1597.94 (10.9%)

A 2 (Group) x 3 (Letter) ANOVA was performed on the averaged data and
individual ANOVAs were carried out when necessary. No main effect of Group was
found (F<1). Again, the letter being searched for influenced the reaction times (F
(2,
10)
=17.22, p=.001). Searching for an N was faster than searching for an M (p<.005).
Nonetheless, W was now faster than in the previous experiment and the reaction time
associated with it was in between that of the M and the N. We checked whether this
pattern of results was due to differential response times for MA and the control group
and found no interaction between them (F
(2, 10)
=1.49, p=.270). Graph 2 shows that, as
93
control participants, MA was faster when looking for a shape made out of N than any
other condition. Even more, as shown by planned comparisons, her reaction times for
M and W were both different from those for N (F
(1, 5)
=8.01, p=.037 and F
(1, 5)
=12.36,
p=.017 respectively). Therefore, the trend for W to be faster than M was surprisingly
due to the control participants and not to MA (F
(1, 5)
=83.69, p=.001).
1000
1500
2000
2500
M N W
S
e
a
r
c
h
i
n
g

R
T
s

(
m
s
)
Controls MA

Again, reaction time data does not point to a clear influence of color aiding MAs
search for an embedded rectangular shape. She only finds a benefit when looking for a
perceptually different display (i.e. N) even though it is supposed to be eliciting the
same color as one of the distracters (i.e. yellow). It seems as though shape and not
color is aiding her performance.
Accuracy analysis
The same 2 (Group) x 3 (Letter) ANOVA was carried out on the accuracy
measures. The groups were now different (F
(1, 5)
=10.96, p=.021 and MA was more error
prone than the control group. An interesting pattern of results for the main effect of
Letter emerged from the data (F
(2, 10)
=10.90, p=.003). Searching for an M display
produced far more errors than any other condition. This pattern could only be due to a
detriment in MAs performance since letter M was categorized for her as the most
difficult to find because it had the same shape as W and the same color as N.
Although the interaction was not reliable (F
(2, 10)
=2.53, p=.129) post hoc tests showed
that control participants did not differ in their error rates for the different letters (all
ps>.7) whereas MA did (see Graph 4). Her error rate for letter M was different than
that for N (p=.032) and that for W (p=.018).
94
0%
10%
20%
30%
40%
M N W
E
r
r
o
r

R
a
t
e
s

(
%
)
Controls MA

Graph 4. Experiment 2. Error percentages for MA and the Control group as a function of
the letter that formed the target stimulus. Error bars denote standard errors of the mean.
This new experiment showed that MA still does not seem to obtain much benefit
from the colors elicited by her synesthesia. In spite of knowing which letter/color to look
for, her performance does not change considerably in relation to the previous
experiment where the letter forming the target shape was unknown until the display
was presented. Especially W keeps proving to be difficult to find or at least, the
benefit expected from its color does not seem to be larger than the benefit found in N
due to its shape. The only hint of a differential pattern for MA is that of the accuracy
measures where her performance for the letter M was worse than for the other letters.
What we found was quite different from what we expected. Instead of being better for
W than for M or N, she was worse for M than for any other letter. It seems as
though she anticipated the difficulty of the task and that influenced her response
accuracy; her synesthesia only seems to have an effect on her search by adding more
heterogeneity to the distracters in the M target condition. Furthermore, she does not
seem to obtain much benefit overall from the previous cue.
In order to further study this, another ANOVA was carried out including now
Experiment as a between block condition. We then had a mixed 2 (Experiment) x 2
(Group) x 3 (Letter) analysis. We found a main effect of Experiment (F
(1, 8)
=58.38,
p=.001) consisting on faster reaction times for Experiment 2. The effect of Letter was
also modulated by Experiment (F
(2, 16)
=6.77, p=.007). Presenting a cue aided
significantly more to the search of rectangles formed by M or W than for those
formed by N. However this benefit was only marginally differential for MA (F
(2, 16)
=3.48,
p=.056). This trend was due to her long reaction times for M and W in Experiment 1
since her reaction times in Experiment 2 were very similar to those of the control group.
95
The same analysis for the accuracy measures showed that there was no
difference between error rates in Experiment 1 and 2. We found that searching for the
different letters did not yield the same results for MA and the Control group although
this was just a trend (F
(2, 16)
=2.82, p=.089). In particular, MAs performance for letter M
was clearly worse than that of Controls for any letter and from her own performance for
N and W (all ps<.03). Even though the second order interaction by experiment was
not significant (F<1), since we expected an influence of cue in MAs performance we
carried out a separate analysis for her data across experiments. In experiment 1 she
was marginally more accurate than in Experiment 2 (F
(1, 8)
=4.66, p=.063) and again
searching for a letter M was more difficult than any other letter (F
(2, 16)
=6.60, p=.008).
The relative effect of letter did not change across experiments (F
(2, 16)
=1.07, p=.365).
This new analysis pours out some remarkable results. Contrary to what would
be expected, MA did not benefit from having a searching cue. Even though her reaction
times decreased, so tended to do her accuracy. Even more, if the cue had any
differential effect on the letters, it was a detrimental one since it only generated an even
less accurate performance for letter M.
Given that the only differences found so far between MA and the control group
were those coming from the accuracy data, we carried out another experiment in which
we promoted differences in this measure instead of reaction time. In order to do so we
showed participants the same display of stimuli for a brief duration of time and
compared their accuracy in discriminating the direction of the rectangle.
EXPERIMENTS 3a and 3b. VISUAL SEARCH WITH SEARCHING CUES AND
LIMITED EXPOSURE TIME
Experiment 3 tried to look for differences between MAs performance and that of
the control group and in order to do so we designed what we thought would be the
ideal circumstances to obtain such differential effects.
Since reaction time showed a great deal of variability among participants, we
modified the task in order to measure accuracy. We showed the display containing the
target rectangle for a limited amount of time and measured correct identification of the
targets direction. Two display lengths were used to increase the chances of finding
differences between the control group and MA in at least one of them.
96
METHOD
Participants
MA and a different group of 10 participants performed this experiment. Nine of
them were females and their mean age was 20. These participants were different from
those performing the previous two experiments.
The task to be performed was the exact same one as in the previous
experiment with the only difference that the searching display was presented for a
limited duration (1000ms or 1500ms). Display duration was manipulated between
experiments. Participants first run the 1000ms display duration experiment.
In each experiment (display duration), six blocks of trials were presented and 24
trials were included in each block (one per location, position and letter condition).
computer running E-prime software (Schneider, Eschman and Zuccolotto, 2002).
Response times were collected with a computer keyboard. A 14 computer screen was
used and auditory feedback was given by means of headphones.
Trials were anticipations had occurred (RT=0ms) were excluded from further
analysis (0.75%). Again, blocks and not participants were used as error factor. Mean
percentage of correct responses per condition can be seen in Table III.
Table III. Experiments 3a and 3b. Error rates per group and experimental condition.
Display Duration
1000ms 1500ms
Letter M N W M N W
MA 42.0% 26.8% 30.4% 37.5% 13.4% 24.1%
CTRLS 35.9% 23.5% 33.7% 25.5% 7.5% 20.2%

A 2 (Display duration) x 2 (Group) x 3 (Letter) ANOVA was performed on the
data and the following results were obtained. An expected main effect of Display
97
duration was found (F
(1, 12)
=15.46, p=.002). More correct responses were given when
the display was presented for longer time on the screen. The main effect of Group was
marginally significant (F
(1, 12)
=4.45, p=.054) and surprisingly, the difference was in the
opposite direction to that expected. As can be seen in Graph 5, Control participants
tended to make fewer mistakes than MA. Last, we found an effect of Letter (F
(2,
24)
=49.64, p=.001). Looking for an N was significantly easier than looking for a W.
Similarly, searching for the W was easier than the M (all ps<.01). Again we checked
whether this pattern was only due to MAs data or the Control participants also showed
it.
0%
10%
20%
30%
40%
50%
M N W M N W
1000 1500
E
r
r
o
r

R
a
t
e
s

(
%
)
Controls MA

Graph 5. Experiment 3a and 3b. Error rates for MA and the Control group as a function of
the time the display was presented and letter that formed the target stimulus. Error bars
denote standard errors of the mean.
A separate ANOVA for the Control group showed a main effect of Letter (F
(2,
24)
=40.63, p=.001) that a further analysis demonstrated to be due to the effect of
searching for an N. This condition yielded more correct responses than any other
(both ps<.001). The same ANOVA carried out for MA showed that the effect of letter
was also significant (F
(2, 24)
=18.30, p=.001). Planned comparisons showed MAs pattern
of results mirrored that found for the main effect of Letter. Searching for a shape made
out of Ns was more precise than searching for shape made out of Ws (F
(1, 12)
=5.33,
p=.040) and the latter more precise than when the shape was made out of Ms (F
(1,
12)
=12.99, p=.004).
Last, the time the display was kept on the screen did not affect the Control
groups results (F
(2, 24)
=1.33, p=.284) nor MAs (F
(2, 24)
=1.03, p=.371). However, planned
comparisons for MA showed that the difference between N and W was only present
98
in the long display condition (F
(1, 12)
=6.00, p=.031) and not in the short one (F<1). M
showed the worst performance (worse than W) in both displays duration (F
(1, 12)
=5.60,
p=.036 and F
(1, 12)
=7.46, p=.018 respectively).
It is therefore clear that MA did not get an obvious benefit from the extra
information coming from the photism and thus it seems that synesthetic colors are not
automatically elicited by the letters and do not automatically guide search in our
synesthete. Only when the conditions were altered so as to maximally promote the use
of that extra dimension did we find a hint of those colors being used. Only when
visualization time was restricted to quite short values did we find that searching for the
letters was modulated by the color. The accuracy when looking for W was not
different from that of letter N. It seems then that only under extreme conditions does
the color (i.e. purple) help find the grapheme (i.e. W). However, the benefit is not as
great as to overcome the facilitation due to the graphemes shape in the case of N. In
relation to M, we found that it was influenced by the previous knowledge about the
identity of the target stimulus. However, this influence goes in an interesting direction.
Knowledge of identity does not benefit performance but hinders it. It seems as though
knowing that the following trial involves dealing with the difficult display does
negatively influence performance.
According to Duncan and Humphreys (1989) two factors play a role in visual
search producing a continuum of search efficiency. As similarity increases between
targets and non-targets the search task becomes more difficult. Likewise, as similarity
between non-targets decreases, the task also becomes more difficult. This theory
would propose that searching for an M would be the most difficult condition since the
target shared photism with one of the distracters and perceptual shape with the other
distracter. Even more, the distracters did not share photism or perceptual shape. In the
case of the target being a rectangle made out of Ws, it shared form with M but did
not share any feature with the other distracter N. Adding to this, both distracters
shared photism. Last, when the N was the target stimulus, both distracters share
shape between them and were different in shape from the target. Only the photism of
one of the distracters was shared with the target stimulus.
If MA is performing the task in a serial manner and color is becoming apparent
only as she searches through the display and identifies each shape, then color will not
aid her in finding the target but it will influence her way to deal with the distracters.
When searching for an N she can group distracters on the basis of their shape; when
searching for a W she can group distracters on the basis of their color. However,
99
when searching for an M distracters cannot be easily grouped since they are
heterogeneous in their color and their shape.
Therefore, the predictions from this theory fit well the accuracy data we
obtained. Looking for a rectangle made out of Ns was consistently found to be the
easiest task to do. Looking for a rectangle formed by Ws was easier than if the
rectangle was made out of Ms in some cases and as difficult in other cases. Last,
when having to find a rectangle made out of Ms MA consistently showed the poorest
performance. On the sole basis of perceptual similarity and dissimilarity between target
and non-targets, Duncan and Humphreys (1989) would predict that responding to W
was as difficult as responding to M. This was the result found for controls in
Experiments 3a and 3b. If synesthetic similarity comes into place, responding to a W
could become as easy as responding to an N, in the case that photism perception
was as strong and automatic as shape perception. Therefore, the relative influence of
the synesthetic perception that is probably varying as a function of the absence vs.
presence of cue and the display duration, could explain the oscillatory findings for letter
W. Independent of the influence of photism similarity, this theory would always predict
M being very difficult to find, as compared to N, and that is what the data shows for
either reaction time or accuracy measures or both.
Following this theory to explain the data we can conclude that, even though
photisms do not seem to be experienced prior to the conscious perception of the
shape, they do play a role in the visual search task by biasing the similarity between
target and distracters and similarity among distracters themselves. Therefore MA does
not show any evidence for pre-attentive photism experience.
It is true that the size of our display was similar to that used in previous
experiments (Palmeri et al. 2002; Ramachandran & Hubbard, 2001a) but the number of
items was increased, thus rendering a more densely populated area and it could be
argued that this is the reason for an absence of effect. In order to check this we carried
out a control experiment in which we showed MA and a group of controls the display as
MA would see it. That is, with letters colored as MA had previously reported that she
perceives them.
100
EXPERIMENT 4. VISUAL SEARCH WITH COLOR CUES
In this last experiment we wanted to make sure that if colors experienced by MA
in the previous experiments had had a strength comparable to that of real displayed
colors, she would have been able to take advantage of that extra dimension of the
stimulus to aid her search even when the density of stimulus on the screen is superior
to that used in other studies.
METHOD
Participants
MA and a different group of 10 participants performed this experiment. These
participants were different from those performing the previous four experiments.
The task to be performed was the exact same one as in Experiment 1 with the
only difference that the searching displays were now colored according to MAs reports
(see Figure 2 for an example of the colored displays).
Six blocks of trials were presented. All other parameters were kept as in
Experiment 1.

horizontal rectangle made out of Ws ; in the middle one the vertical rectangle is made
out of Ns and in the right one the horizontal rectangle is made out of Ms .
101
A total of 17% of the trials were eliminated from the reaction time analysis due
to incorrect response. Again a 2SD cut off was used to eliminate outliers. The
remaining data were submitted to a 2 (Group) x 3 (Letter) ANOVA. Mean reaction times
and error rates for each condition and participant can be seen in Table IV.
Table IV. Experiment 4. RT (and percentage of errors) per group and experimental
condition.
Letter M N W
MA 882.33 (2.1%) 1087.81 (11.1%) 642.26 (2.1%)
CTRLS 1564.34 (21.0%) 1806.31 (24.9%) 1054.89 (7.8%)

Groups were now different (F
(1, 5)
=86.67, p=.001). MA was faster than control
participants. This may have been due to her extensive practice after having performed
the previous four experiments. We also found a significant main effect of Letter (F
(2,
10)
=180.57, p=.001). As expected, W was faster than M or N (F
(1, 5)
=114.83, p=.001
and F
(1, 5)
=629.67, p=.001 respectively). Surprisingly, a rectangle made out of Ms was
also faster to be found than when it was made out of Ns (F
(1, 5)
=40.15, p=.001). This
finding was unexpected since both letters were presented in the same color. If
anything, we would have expected that due to its unique form, N and not M would
have been easier to find. A significant interaction by group (F
(2, 10)
=18.95, p=.001)
showed that control participants found more benefit from having W colored in purple
than MA did (see Graph 6).
102
400
900
1400
1900
M N W
S
e
a
r
c
h
i
n
g

R
T
s

(
m
s
)
Controls MA

The finding that M yielded faster reaction times than N was surprising as it
was unexpected. It could be that since letter M was perceptually bigger than letter
N, the yellow spot formed by a rectangle made out of Ms was more apparent and
thus easier to categorize as vertical or horizontal without even processing the identity
of the letter forming it. Nevertheless, even though expected, the most interesting result
is the relative benefit of letter W as opposed to the other letters. Finding these results
with MA demonstrates that she is not just bad at searching for items in a display but
rather, when presented with achromatic graphemes, color (photism) does not become
conscious in time to exert a noticeable effect on her search.
Accuracy analysis
The same ANOVA carried out on the error percentage showed again a different
performance for both two groups (F
(1, 5)
=64.22, p=.001). MA was more accurate than
the control participants. Again, this could be due to extensive practice with the task.
Finding a W yielded more accurate responses than any other condition (F
(2, 10)
=18.24,
p=.001). This time the difference between M and N was not reliable (p=.157) and
this was due to a differential error rate pattern for both groups (F
(2, 10)
=5.41, p=.026).
As can be seen in Graph 7, MA was as accurate for M as she was for W
while control participants did not show a difference between yellow colored letters.
Although counterintuitive, she is definitely obtaining a benefit from answering to letter
M over letter N. Although we do not readily have an explanation for MAs results, it
could well be that after having performed the same visual search task so many times,
she had developed some strategies to search for the letters, especially for letter M
103
that was always referred to as the difficult one and such strategies aided her search
when color was present giving rise to a lower error rate for M than N.
0%
10%
20%
30%
M N W
E
r
r
o
r

R
a
t
e
s

(
%
)
Controls MA

Graph 7. Experiment 4. Error rates for MA and the Control group as a function of the
Given that elicited photisms do not seem to automatically guide visual search in
MA, our next concern was to study whether, although experienced after the cognition of
the inducer, they were still strong enough as to influence behavior when the complexity
of the display was raised and most important, whether attention would modulate such
influence in any manner.
EXPERIMENT 5. ATTENTION MODULATION OVER PHOTISMS ELICITED
BY COMPLEX STIMULI
To study how attention modulates perception of photisms elicited by letters we
followed up previous experiments by Rich and Mattingley (2003). There they presented
a group of synesthetes with a set of Navon like displays where the color elicited by the
stimuli forming the global and the local shape could be either congruent or incongruent
with the color in which the display was presented.
As previously mentioned, in the original study they carried out two experiments
to test whether ignoring the dimension of a stimulus that elicited a photism incongruent
with the display color would be efficient enough as to avoid the photisms influence on
performance. They predicted that a photism induced by the letter that was being
attended would compete more with the display color than the photism associated to the
letter in the dimension of the stimulus that was being ignored. In Experiment 1, where
attention was not manipulated, they found that when both local and global letters elicit
104
the same photism and it was congruent with the display color, synesthete participants
were faster to name the display color than when it just matched the photism elicited by
only one of the two dimensions (either the global or the local). Last, the condition where
both dimensions were eliciting a color incongruent with the displayed one was the
slowest. That is, a congruently colored stimulus facilitated performance while an
incongruently colored one worsened performance. When a dimension of the stimulus
was congruent and another dimension incongruent, naming times were in between.
In the second experiment they manipulated, by means of instructions, the
dimension of the stimulus that the person was paying attention to. Now, in each case
the attended dimension was always congruent and the ignored dimension could either
be also congruent or incongruent. They then compared the congruency effect in these
circumstances with that found in Experiment 1 for the same conditions where attention
was not manipulated. Results from Experiment 2 showed that when participants were
requested to attend to the global form and the ignored local form was incongruent, they
were slower than when both dimensions were congruent. This congruency effect was
as large as that found in Experiment 1 for those two conditions. When the instructions
asked to attend to the local form and the ignored global form was incongruent, again
slower naming time was found as compared to the same both-levels-congruent
condition. Here, they found that the congruency effect was reliably smaller than that
found in Experiment 1 for the very same conditions. That is, when attending local,
attention seemed to attenuate the interference coming from a global incongruent color.
Several problems arise from this design. First, they did not make sure their
participants were in fact paying attention to the dimension they had been instructed to.
Second, because of the way they designed their experiment, they were confounding
color congruency with letter consistency. Both level congruent conditions were also
letter-consistent (global and local letters were the same) while one level congruent-one
level incongruent was also letter-inconsistent (global and local letters were different).
Third, in the second experiment only 2 of the original 4 conditions were included.
Therefore they did not check the effect of attending to an incongruently colored
stimulus and ignoring a congruently colored one, or whether attending to one
dimension when both dimensions were incongruent would produce a smaller
congruency effect. Fourth, the critical comparison to check the role of attention was
made across experiments. Congruency effect when participants were supposedly
attending to one specific dimension was compared to the congruency effect found in
experiment one for the same conditions. However, there is no evidence whatsoever as
105
to what was being attended in Experiment 1. The logical comparison would have been
that of the different original conditions when they were attended vs. ignored. Last, they
tested and averaged performance from 17 synesthetes of whom no information is
provided regarding the type of synesthesia experienced (i.e. higher vs. lower / projector
vs. associator) and this renders their results very difficult to interpret.
The aim of our experiment was to complement this study and also to test how a
higher synesthete would behave under such circumstances. Even though the Stroop-
NP experiment showed that her colors were automatically elicited and difficult to ignore,
the Visual Search experiments showed that they were not strong-vivid enough as to
pop out and therefore guide and make easy a difficult visual search task.
METHOD
Participants.
MA and a group of 10 participants carried out the task. They all had normal or
corrected to normal vision and the mean age of the group was 22. Eight of them were
women (including MA).
Design and Procedure.
The task to be performed was to name the display color of a set of stimuli
appearing on the computer screen while ignoring their identity. Stimuli were letter
shapes make out of small colored letters.
The variables being manipulated were 3: Attention (to the global form or the
local form), Letter Consistency (whether the global and local letters were the same or
different) and Photism-Color Congruency (whether the color elicited by the stimulus
was congruent or incongruent with the display color).
Four blocks of trials were presented. In two of them participants had to pay
attention to the global form and in the remaining two they had to attend to the local
form. The order of the blocks was counterbalanced across participants. MA run the
experiment twice, so that she went through both counterbalancing conditions.
Four different colors were used and two stimuli of each color were presented.
The stimuli and their photisms are shown in Table V. With this design we could
manipulate Letter Consistency as we kept Photism-Color Congruency constant by
106
presenting two letters eliciting the same photism, one in the global level and another
one in the local level. Also, this set of stimuli enabled us to have pure Photism-Color
Congruency conditions (i.e. both levels congruent) even when the Letter display was
inconsistent. Therefore Photism-Color Congruency had 2 levels for the consistent letter
level (i.e. 2 pure levels: a B made out of Bs in red congruent- or in blue
incongruent-) and 4 levels for the inconsistent letter level (2 pure levels where both
letters elicited the same photism: congruent a red B made out of Es- and
incongruent a blue B made out of Es and 2 mixed levels were the color only
matched the photism of one dimension of the stimulus: only global letter congruent a
red B made out of Os- and only global letter incongruent a yellow B made out of
Ms).
Table V. Experiment 5. Experimental design and items used with the corresponding
elicited photism. Only the Attend Global condition is shown since the stimuli were the
same and only instructions to attend to the global or local form varied between
conditions.
L S K Z B E M O
Attention Global
Letter
Consistency
Consistent Inconsistent
Pure Cong: Pure Incong: Pure Cong: Mixed Incong: Mixed Incong: Pure Incong:
Photism-
Color
Congruency

G-cong
L- cong
G-incong
L-incong

G-cong
L- cong
G-cong
L-incong
G-incong
L- cong
G-incong
L-incong

In order to ensure that the Attention manipulation was effective, 11% of trials
were catch trials where participants had to detect the presence of a letter and report its
identity but only if that letter was presented in the dimension of the stimulus that was
being attended (i.e. in a attend global condition, they had to detect the presence of a
letter T only if this was the global form made out by other letters but not if it was the
local letter making up a global form). Therefore, in half the catch trials participants had
to report the identity of the letter (when it was presented at the level they were
attending) and in the other half they had to report the color of the stimulus (when it was
presented at the level they were supposed to be ignoring). Letter T was chosen
because it elicits a black photism for MA, and therefore would not interfere with the rest
107
of the conditions. Attention was manipulated between blocks, and Letter Consistency
and Photism-Color Congruency within blocks.
For each level of Letter Consistency, each block had equal number of trials for
each pure level of Photism-Color Congruency (24 per level) and a slightly smaller
number of mixed-level trials (16 per level). Therefore out of the 144 trials per block 48
were letter-consistent, 80 were letter-inconsistent and 16 were catch trials. All trial
types were randomly presented so that participants could not anticipate them.
In each trial a fixation point (a + in the middle of the screen) was presented for
750ms. Afterwards, the target stimulus was shown and remained on the screen until
participants named the color to the microphone. Once they had given a response the
experimenter recorded it on the keyboard for accuracy measurement purposes and the
next trial began.
Stimuli and material.
At a viewing distance of about 53cm global letters were from 4.32 to 6.48 wide
and from 6.48 to 7.02 tall. Local letters were 0.65 wide and 0.65 tall. Global letters
were made by a variable number of local letters (from 12 to 25) depending on their
structure (i.e. letter L was made of 12 letters while letter M was made of 25 letters).
computer running E-prime software (Schneider, Eschman & Zuccolotto, 2002). Verbal
naming times were collected with a voice key microphone connected to a serial
response box. A 14 computer screen was used.
Catch trials were analyzed to make sure that participants were following
instructions. All participants had performances well above chance (lowest individual
performance on a category: 81%). Therefore all participants were included in the
statistical analysis
1
. Trials were the microphone had been activated due to extraneous

1
Six out of the 10 participants had lower accuracy scores in the catch trials than MA. A subsequent analysis
including only the remaining 4 participants was carried out and none of the results differed from those shown here.
Therefore those analyses are not reported.
108
noise or it had failed to detect a response (2.75%) were excluded from the analysis. A
further 1.2% was excluded due to incorrect response. Naming times faster than 200ms
or slower than 1200ms were considered outliers and therefore also eliminated from
analysis (2.5%).
The mean for MAs scores for each item over the 2 sessions were averaged as
so were the scores for the ten participants. Therefore, items and not subjects were
used as the error factor in the subsequent ANOVAs. Two statistical analyses were
carried out to analyze the data. We first analyzed the pure-level conditions and in order
to do so we carried out a repeated measures ANOVA with Group (2) x Attention (2) x
Letter Consistency (2) x Photism-Color Congruency (2) as within-item factors. When
second or third order interactions were significant, separate ANOVAs were carried out
for each group to further study them. Reaction times for each experimental condition
can be found in Table VI.
A main effect of Group showed that MA was slower than the control participants
(F
(1, 7)
=283.55, p=.001). Also, the effect of Attention was significant in the sense that
when participants attended to the global form they were faster than when attending to
the local form (F
(1, 7)
=9.12, p=.019). No main effect of Letter Consistency or Photism-
Color Congruency was found (F
(1, 7)
=1.5, p=.260 and F
(1, 7)
Table VI. Experiment 5. RT per group and experimental condition.

Attention Global Local
Letter
Consiste
ncy
Consistent Inconsistent Consistent Inconsistent
Pure
Cong
Pure
Incng

Pure
Cong
Mixed
Incng
Mixed
Incng
Pure
Incng

Pure
Cong
Pure
Incng

Pure
Cong
Mixed
Incng
Mixed
Incng
Pure
Incng
Photism-
Color
Congrue
ncy
G-
cong
L-
cong
G-
incng
L-
incng

G-
cong
L-
cong
G-
cong
L-
incng
G-
incng
L-
cong
G-
incng
L-
incng

G-
cong
L-
cong
G-
incng
L-
incng

G-
cong
L-
cong
G-
cong
L-
incng
G-
incng
L-
cong
G-
incng
L-
incng
MA 772.33 807.75 757.04 878.06 849.96 878.59 805.37 804.23 777.22 894.40 826.06 846.18

Ctrls 663.51 649.42 691.47 698.46 700.37 657.42 703.46 686.26 719.42 743.78 719.08 703.34

Group x Letter Consistency was not significant either (F<1) so it seemed as
though the inconsistency between global and local letter did not affect MA differentially.
Nevertheless, Photism-Color Congruency did produce different results for each group
in the expected direction (F
(1, 7)
=56.96, p=.001). Responding to congruent trials was
faster than responding to incongruent trials, but only for MA, not for the control
109
participants (F
(1, 7)
=8.96, p=.020 vs. F<1). This tells us that MA was processing the
photism elicited by the stimulus, in such a way that it interfered with her color naming
task.
Photism-Color Congruency was also modulated by Attention (F
(1, 7)
=23.52,
p=.002). The congruency effect was much larger when participants had to attend global
than local. The second order interaction by group was marginally significant (F
(1,
7)
=4.10, p=.083) and separate ANOVAs for MA and our control group showed that this
interaction was due to MAs pattern of responses, which mirrored the general one (F
(1,
7)
=11.66, p=.011). Control participants showed no effect (F<1). While control
participants were faster when attending global than local independently of Photism-
Color Congruency, MA showed a more complex pattern. The congruency effect was
present when attending global (p<.003) but not when attending local (p=.127). This is
interesting because the conditions included in this analysis were those where either
global and local letter were the same and thus both levels were congruent or
incongruent, or global and local letters were different but elicited the same photism, so
that were colored either congruently with both letters or incongruently with both letters.
Therefore, attending global did not provide with information different from that obtained
when attending local. We can then only suggest that when MA is attending local she
does not process letter identity to the level of photism activation and therefore obtains
no benefit/interference from that information.
550
650
750
850
950
C
o
n
s
i
s
t
e
n
t
I
n
c
o
n
s
i
s
t
e
n
t
C
o
n
s
i
s
t
e
n
t
I
n
c
o
n
s
i
s
t
e
n
t
C
o
n
s
i
s
t
e
n
t
I
n
c
o
n
s
i
s
t
e
n
t
C
o
n
s
i
s
t
e
n
t
I
n
c
o
n
s
i
s
t
e
n
t
Global Local Global Local
Controls MA
N
a
m
i
n
g

T
i
m
e
s

(
m
s
)Congruent Incongruent

Graph 8. Experiment 5. Reaction times for MA and the Control group when responding to
pure-level conditions. Error bars denote standard errors of the mean
110
We also found a significant second order interaction between Group, Letter
Consistency and Photism-Color Congruency (F
(1, 7)
=8.26, p=.024). In the separate
ANOVA for each group we found no effect for the control participants (F<1) and a
significant interaction between Letter Consistency and Photism-Color Congruency for
MA (F
(1, 7)
=7.65, p=.028). Graph 8 shows the results found. We predicted a permanent
presence of a congruency effect that could be enlarged in those cases where the letter
was inconsistent since those are the conditions where color could help or hinder
performance. However, what we found was no Photism-Color Congruency effect when
letters were consistent (i.e. a B made out of Bs) (F<1). Congruently colored displays
were named as fast as incongruently colored displays. When the letter was not
consistent, we did find a congruency effect (F
(1, 7)
=27.27, p=.001). Responses to
incongruent trials were slower than responses to congruent ones. Again, this pattern of
results was unexpected. Photism-Color congruency only affected naming times when
the two photism eliciting stimuli were different (i.e. different global and local letter).
Overall, in this first analysis we found that MAs Photism-Color Congruency
effect was influenced by Attention (only when attending global was the congruency
effect found) and by Letter Consistency (only when letters were inconsistent was the
congruency effect found). This data might suggest that attending local was more
difficult than attending global (as stated by the Global precedence effect) and therefore
more resources were devoted to that task and the experience of synesthesia was
somewhat diminished. It could also be that the stimuli used in this task were so
complex that only when letter identity processing was promoted (i.e. attending to the
global form or having an inconsistent stimulus) did MA process it to the extent as to
activate the photism associated with it.
In the second analysis we were interested in the influence of Photism-Color
Congruency as a function of the dimension were attention had been allocated. We
therefore analyzed those trials were the Photism-Color Congruency was present only
at the global level or only at the local level and compared them with those trials were
both levels were congruent or both levels were incongruent. As mentioned before, in
order to have these Photism-Color Congruency levels, only inconsistent letter
conditions could be used for the analysis. The ANOVA had the following factors: Group
(MA and Ctrl) x Attention (Local and Global) x Global Photism-Color Congruency
(global level congruent and global level incongruent) x Local Photism-Color
Congruency (local level congruent and local level incongruent).
111
Again we found that MA was overall slower to respond than control participants
(F
(1, 7)
=178.34, p=.001). Attending global produced faster RTs than attending local (F
(1,
7)
=5.90, p=.046) and this was so for the control group but not for MA (significant Group
x Attention interaction, F
(1, 7)
=14.08, p=.007). No main effect of Global Photism-Color
Congruency or Local Photism-Color Congruency was found. We expected that these
variables would be influenced by Group but it was so only for the Group x Local
Photism-Color Congruency interaction (F
(1, 7)
=8.65, p=.022) but not for the Group x
Global Photism-Color Congruency (F
(1, 7)
=2.49, p=.158).
A separate ANOVA for each group showed that control participants had no effect
of either variable (both Fs<1) while MA had an effect of Local Photism-Color
Congruency (F
(1, 7)
=5.98, p=.044) but no effect for the Global Photism-Color
Congruency (F<1). Here her pattern of results was not different from that of the control
participants: congruent and incongruent trials had similar reaction times. For the Local
Photism-Color Congruency she was faster on locally congruent trials than on locally
incongruent ones.
We also expected that if attention played a role in synethesia it would modulate
the Photism-Color Congruency effect for MA but not for the control participants so that
attending global would attenuate the local incongruence and attending local would
diminish the global incongruence. None of these two second order interactions were
significant (F(1, 7)=1.66, p=.239 for Global Photism-Color Congruency and F(1,
7)=1.07, p=.335 Local Photism-Color Congruency). A subsequent ANOVA for each
group showed that for control participants the Local Photism-Color Congruency was not
affected by the dimension to which attention was being paid (F(1, 7)=2.91, p=.132, see
Graph 9a left side). Likewise, for Global Photism-color Congruency controls did not
differ in their naming times as a function of attention (F<1, Graph 9b left side). For MA,
the local congruency effect was the same when she was attending local or global form
(F<1). As clearly shown in the Graph 9a, attention did not modulate the congruency
effect coming from the local dimension. Regarding the congruency in the global
dimension, although a congruency effect was seen when attention was paid to the
global form and not when attention was paid to the local form (see Graph 9b right side),
the Attention x Global Photism-Color Congruency interaction did not reach significance
(F(1, 7)=2.47, p=.159)
112
550
650
750
850
950
Controls MA
N
a
m
i
n
g

T
i
m
e
s

(
m
s
)
L-Congruent L-Incongruent

550
650
750
850
950
Controls MA
N
a
m
i
n
g

T
i
m
e
s

(
m
s
)
G-Congruent G-Incongruent

Graph 9a and 9b. Experiment 5. Reaction times for MA and the Control group when
responding to inconsistent-letter conditions. a) Conditions where the photism of the
local letter was incongruent with the displayed color (collapsed for global letter
congruency). b) Conditions where the photism of the global letter was incongruent with
the displayed color (collapsed for local letter congruency). Error bars represent standard
errors of the mean.
Since the variability of the data was too large to find statistically significant
effects, we collapsed Photism-Color Congruency levels as a function of whether
attention was being paid to the congruent or the incongruent level. Therefore, attending
local levels where the Local Photism-Color Congruency was congruent were collapsed
with those trials were the participant had to attend global and the Global Photism-Color
Congruency was congruent. The same process was carried out for those conditions
where attention was paid to the incongruent level. Last, these 2 resulting conditions
were compared to those where either attended and ignored level were congruent or
both attended and ignored levels were incongruent. We therefore compared MA and
the Control Group for the 4 levels of Attended-Photism-Color Congruency (attended-
congruent, attended-incongruent, both levels-congruent and both levels-incongruent).
113
This new 2 (Group) x 4 (Attended-Photism-Color Congruency) ANOVA showed
that the interaction between Group and Congruency was statistically significant (F
(1,
7)
=12.24, p=.001). Control participants were not different independently of the Attention-
Congruency condition (all ps>.3). However, MA showed a very different pattern of
results. As expected, both levels-congruent trials were faster than any other type of trial
(all ps<.001). Nonetheless, as can be seen in Graph 10, an interesting result was found
for the remaining 3 conditions. MA was no different when the attended level was
incongruent than when both levels were incongruent (p>.9). Even more, those 2
conditions were not different from trials where only the ignored level was incongruent
(all ps>.8).
550
650
750
850
950
Both levels Attended level Both levels Attended level
Controls MA
N
a
m
i
n
g

T
i
m
e
s

(
m
s
)

congruency of the attended level. Error bars denote standard errors of the mean.
This very interesting pattern of results can be interpreted in the light of another
significant interaction found in the previous analysis (see Graph 11). An unexpected
interaction between Global Photism-Color Congruency and Local Photism-Color
Congruency was found ((F
(1, 7)
=25.88, p=.001). The congruency effect for each variable
was only found on the congruent level of the other variable (i.e. Local Photism-Color
Congruency effect was only found when the Global color was congruent and vice
versa). Although only marginally significant (F
(1, 7)
=4.14, p=.082), this pattern of results
tended to be found in MA and not in control group.
114
550
650
750
850
950
G-Congruent G-Incongruent G-Congruent G-Incongruent
Controls MA
N
a
m
i
n
g

T
i
m
e
s

(
m
s
)
L-Congruent L-Incongruent

Global and Local Photism-Color Congruency. Error bars denote standard errors of the
mean.
A separate analysis confirmed that MAs results followed this pattern (F
(1,
7)
=16.69, p=.005) while control participants did not. This interaction could explain the
fact that attending to the congruently colored dimension of a stimulus would not
produce faster reaction times than attending to the incongruently colored dimension. It
seems as though when one dimension is already incongruent, the fit of the other
dimension to the internal expectations is not computed. Stimuli are either right or wrong
but not half way through. That is, MA finds no benefit from attending to the congruent
level and ignoring the incongruent level.
Almost no errors were made in this experiment, with, many experimental
conditions having a 0% error rate. Therefore no statistical analysis of accuracy was
carried out.
GENERAL DISCUSSION
The first series of experiments were aimed at studying whether MAs photisms
are elicited pre or post attentively. Results from the first four experiments consistently
show that MA does not seem to find any benefit from her internally elicited colors when
searching for an embedded figure. That is, letter induced photisms do not seem to pop
out the same way as real color do in standard perception when searching for an
embedded figure. This conclusion is backed up by Experiment 4 where real colors
were presented and MA showed a fairly different pattern of data. Similar results have
been found by other groups when comparing synesthetes performance in achromatic
115
visual search tasks with controls performance when shown chromatic stimuli (Hubbard
et al., 2005).
Importantly although the photism elicited by a figure when it is made out of Ws
does not pop out, the photisms induced by the letters do have an influence on MAs
search. When searching for a figure made out of Ms MA was overall slower and made
more errors than controls and her when searching for a figure made out of Ws. Note
that in this condition the target letter shares color with one of the distracters (N) and
shape with the other distracter (W). Even more, distracters are dissimilar in this
condition, as they induce different photisms. This increased target-distracter similarity
and distracter heterogeneity does explain why this condition is the worst for MA
(Duncan & Humphreys, 1989). Thus, although photisms are not automatically elicited,
at least at low level perceptual levels (they do not pop out), they are automatically
elicited in the sense that they seem to be activated in spite of being detrimental to
performance. That is, from the visual search experiments we can conclude that
photisms do not seem to be easily inhibited, which is in clear agreement with the
results from the Stroop-NP experiments (Experiments 2a and 2b in Chapter 1).
Experiment 5 on the other hand, was aimed at studying the influence that
selective attention can exert over photisms and their effect on performance. Results
from the second analysis show that the congruency effect produced by the perception
of a stimulus not colored according to MAs schema is quite strong. Even more, it
seems to be independent of the dimension of the stimulus to which attention is being
paid. As long as one dimension of the stimulus (either the local or the global form) is
incongruently colored MA performs as bad as when both dimensions are wrongly
colored. Even more, attending to the incongruently colored dimension and ignoring the
congruently colored dimension yields the same results as attending to the congruently
colored one and ignoring the incongruently colored one. These results are similar to
those found in the Stroop-NP experiment (Lupiez & Callejas, 2006) reported in
Chapter 1 where the interference produced by photisms on the color naming task was
much larger than that produced by the wavelength of the video display on the photism
naming task. We then suggested that the emotional response associated to the
perception of an incongruently colored stimulus could be a possible explanation of the
results found. Here again, MA behaved atypically while performing this task. She
reported experiencing a high level of anxiety and discomfort. The emotional state
derived from giving answers contrary to what would be natural for her (that is, inhibiting
naming the photism in favor of the video color) affects her in a way that prevents
116
normal focusing of attention from working properly. Callejas, Acosta and Lupiez
(submitted, Chapter 3) have shown that in fact affective reactions can be evoked by the
incongruence between external color and internal photism and they can influence
behavior.
It could be argued that the attentional manipulation might have not been
effective and that is why MA did not show a differential pattern between those trials
were she was attending to the congruent dimension and those where she was
attending to the incongruent one. An analysis of catch trials showed that MA, as well as
control participants, performed the secondary task with a level of accuracy that indexes
that instructions were obeyed. Therefore we do have grounds to state that while
attention was being paid to one dimension, the incongruence at the ignored dimension
was affecting naming times.
Our results from this last experiment are in disagreement with those found by
Rich and Mattingley (2003) where they found a modulation of the congruency effect
when participants were asked to attend local and the global dimension was
incongruently colored. Any of the methodological issues outlined in the introduction
could account for these inconsistencies. We lean towards two of them being the most
influential ones. On the one hand, they did not have a way to check whether
participants were actually following instructions and paying attention to the requested
dimension. Note that they limited their analysis to compare a condition where attention
was not manipulated, and therefore they did not know how attention was working, with
a condition from a separate experiment where they assumed that attention was being
paid to the requested dimension. On the other hand, synesthete data was coming from
an average of fourteen subjects and it would be extremely improbable that all fourteen
showed the pattern reported in their paper.
Taken together, these findings highlight the importance of taking into account
individual differences when arriving to conclusions and also the need to study other
factors that could be influencing the synesthetes performance. Ramachandran and
Hubbard (2001b) and Smilek and Dixon (2002) pointed out to the difference between
higher and lower synesthetes or associator vs. projector synesthetes and later work
has shown empirically that this labels do actually represent groups of people that
behave differently when confronted with the same task (Dixon et al., 2004). Hubbard et
al. (2005) suggests, based on his behavioral and neuroimaging correlations, that there
might be even more subcategories of synesthetes. It might be interesting and a good
predictor of synesthetes performance to know the extent to which perceiving a stimulus
117
in the wrong color affects them emotionally. We now turn to study this topic in our
synesthetes.

SECTION 2:
SYNESTHESIA AND EMOTION

CHAPTER 3
GREEN LOVE IS UGLY: EMOTIONS ELICITED BY
SYNESTHETIC GRAPHEME-COLOR PERCEPTIONS
1

1
This chapter has been submitted as a Research Article to Journal of Cognitive Neuroscience being its authors Alicia
Callejas, Alberto Acosta and Juan Lupiez.
Chapter 3. Green Love Is Ugly _______________________________________________
123
ABSTRACT
Synesthetes experiencing grapheme-color synesthesia often report feeling
uneasy when dealing with incongruently colored graphemes, although this issue has
not been thoroughly investigated. We experimentally studied these affective reactions
related to synesthetic perceptions by means of an evaluation task. We found that the
perception of a wrongly colored word affects the judgments of emotional valence.
Furthermore, this effect competed with the words emotional valence in a categorization
task: responses were slower (and less accurate) for the inconsistent conditions (i.e.
when positive-valence words were presented in a color different from the synesthetes
photism -producing a negative affect- or when negative-valence words were presented
in the photisms color-producing a positive affect-) as compared to the consistent
conditions, thus supporting the automatic nature of these synestheticaly elicited
affective reactions. Therefore, the congruency effect that has been taken as an index of
the true experience of synesthesia (i.e. faster reaction times to congruently colored
words than incongruently colored words) can be reversed when the experimental
manipulations turn an incongruently colored word into a more consistent stimulus by
aligning the result of the affective reaction produced by the word-color congruence with
the valence assessment of such word. This is the first time that an affective reaction
elicited by the congruency between the synestheticaly induced color of a word and the
color in which the word is actually presented is reported. The underlying neural
mechanisms that might be involved in this phenomenon are discussed.
INTRODUCTION
Synesthesia is known as a mixing of the senses in which stimuli from one
sensory modality induce experiences proper of a different modality (Beeli, Esslen &
J ncke, 2005; Cytowic, 2002; Dixon, Smilek, Caudahy & Merikle, 2000; Mattingley,
Rich, Yelland & Bradshaw, 2001; Robertson, 2001). Thus, taste can be experienced as
a shape (Cytowic, 1998), or sounds can are experienced as specific tastes (Beeli,
Esslen & J ancke, 2005). The most common form however, involves mixing two
features from the same visual modality, i.e., colors (called photisms) subjectively
experienced when perceiving non-colored alphanumerical characters (Day, 2005).
Different theories have been put forward to account for this unusual phenomenon being
one of them the promoter of the idea of a cross-wiring between areas of the brain
involved in grapheme form processing and color processing (Ramachandran &
________________________________ SECTION 2. SYNESTHESIA AND EMOTIONS
124
Hubbard, 2001b). Since the path for word processing and color processing seems to
be somewhat parallel in terms of neural substrates, depending on where the cross-
wiring takes place, different types of grapheme-color synesthesia could be accounted
for (Ramachandran & Hubbard, 2001b). These authors suggest that a crosstalk taking
place in the fusiform gyrus (where grapheme form is processed as well as color-
V8/hV4-), would be the mechanism underlying lower synesthesia -a form of grapheme-
color synesthesia that is of a more perceptual nature-. If the crosstalk takes place at
higher instances of processing such as the angular gyrus where the next stage in color
processing takes place as well as the abstract numerical calculation, then the resulting
phenomenon would be what is called higher synesthesia where associations between
colors and graphemes seem to be of a more conceptual level (i.e., driven by the ordinal
aspect of the graphemes). They have recently provided evidence that hV4 is
differentially active in synestehtes but not controls when looking at synesthesia
inducing stimuli (i.e. graphemes) (Hubbard, Arman, Ramachandran & Boynton, 2005).
This theory could also account for the fact that a normal individual could
experience synesthesia-like phenomena when under the influence of such drugs as
LSD if we assume that the cross-activation is due to existing connections that are only
active in synesthetes and could become active in normal individuals when taking these
substances although these authors favor the idea that these connections are only
found in synesthetes and that synesthesia like phenomena experienced when
consuming drugs is qualitatively different (Hubbard & Ramachandran, 2005).
A commonly known but poorly studied feature of grapheme-color synesthesia is
the experience of emotions associated to the synesthetic event. There is a sense of
certitude that somehow conveys the feeling that ones experiences are the right ones
(Cytowic, 1993). More specifically, synesthetes experience a positive feeling when they
see a consistent stimulus (i.e. a letter colored according to his/her photism for it in
grapheme-color synesthesia) and a feeling of discomfort associated to the experience
of a mismatch between the physical-external stimuli and the subjective-internal
synesthetic perception (i.e. a letter not colored according to his/her photism for it). It
has been common in the synesthesia literature to find informal reports about negative
emotional reactions associated to the experience of stimuli presented in a color
different from their photism. If a synesthete who sees the word tree in blue is
presented with such word colored in pink, he/she will say This is wrong! and this
reaction can vary in intensity from a discomfort statement to a fairly strong aversive
emotional response.
125
Even though this affective reaction is easily found among synesthetes, to our
knowledge no empirical study has been carried out to test whether it is in fact a
measurable and reliable phenomenon that influences synesthetes behavior. Studies
have been undertaken to elucidate the possibility that emotions are the inducer of
synesthetic experiences (Ward, 2004) so that only words that have an emotional
connotation elicit a synesthetic response. Nevertheless, no study has been carried out
on the emotions elicited as side effect of the perception of an incongruently colored
stimulus in grapheme-color synesthesia.
As Ramachandran and Hubbard suggest (2001b), the fact that synesthetic
perceptions could in turn produce emotional reactions is highly interesting from a
neuroscientific viewpoint and also as another piece of evidence in favor of their cross-
wiring theory since it is known that information processed in the temporal lobe gets
relayed to the amygdala and other parts of the limbic system (Amaral, Prince, Pitanen
& Carmichael, 1992; LeDoux, 1992).
Therefore, we wanted to study the nature of these affective reactions. Following
the logic established in previous literature to differentiate developmental synesthesia
from other experiences such as mere associations or metaphors (Cytowic, 2002; Dixon
et al., 2000; Ramachandran & Hubbard, 2001b), our first aim was to determine whether
these reactions are just affective memory associations or they are automatically elicited
by the perception of an incongruently colored grapheme as photisms themselves are.
In order to do so, we investigated synesthetic perception of emotional words in
MA, a Psychology undergraduate student that experiences grapheme-color
synesthesia (Lupiez & Callejas, 2006). She reports experiencing negative emotions
associated to the perception of letters, numbers and words when presented in a color
different from her photism for them. She describes her feelings by saying It is wrong,
its like coming into a room and finding all the chairs upside-down and everything out of
place. I cant stand it. It is just wrong. When performing a Stroop experiment to
evaluate the automaticity of her synesthesia (i.e., naming the colors of letters and
numbers that elicited either the same or different colors as her photisms for them) her
discomfort could be overtly perceived (rapid increase of hand perspiration, difficulty to
sit still, constant posture readjustments, etc). In this paper we report four experiments
where we studied the behavioral influences of these affective reactions.
126
EXPERIMENTS 1a AND 1b. VALENCE JUDGMENT TASK
Our first aim was to try to reproduce the reported disliking with an indirect
behavioral task. Since MA perceives words as clearly and uniformly colored (as
opposed to perceiving one color for each letter), we presented her with a set of words
and asked her to rate their valence according to their semantic meaning. In order to
test the influence of the synesthetic colors, we manipulated the color in which the
words were presented so that it would either match (congruent condition) or mismatch
(incongruent condition) that of the synesthetic experience. This was compared to an
absence of color condition where all the words were presented in black. Informal
reports by synesthetes points to black being considered as a neutral color maybe due
to the enormous amounts of practice acquired in reading texts in black ink.
METHOD
Participants
Our synesthete MA and a group of 11 control participants took part in this
experiment. Participants matched MA for age (22) and gender and they were all
students at the Psychology Department at the University of Granada. MA ran the
experiment in two different occasions while control participants ran it only once.
Experiment 1a. Evaluation task
Design
Two variables were manipulated in this experiment: word valence and color
congruency. To manipulate word valence a set of 72 words were used for the
experiment. There was an equal number of neutral words, positive words, anger-
related negative words and anxiety-related negative words, being the four groups of
words matched in length and subjective familiarity. The words were chosen from a
database of words used in previous experiments in our laboratory so that the mean
rating for positive words was +2, for neutral words 0 and for negative words (anger
related as well as anxiety related) was -2, in a scale going from -3 to +3. Words were
presented both in congruent and incongruent color. Congruent colors were those that
MA had previously chosen for each word as part of a study to check consistency over
time. Incongruent colors were the color-wheel opposites of each chosen color.
127
Procedure
Randomly ordered words appeared at the centre of the screen one at a time
and participants task was to evaluate the valence of each word without time pressure
in a 7-point Likert scale by means of a keyboard and being +3 very positive, 0 neutral
and -3 very negative. A white fixation point was presented against a gray background
for 500 ms. It consisted on a cross in the middle of the screen and it was accompanied
with a reproduction of the responding scale (i.e. the possible rating values from -3 to
+3) that was also white and appeared at the bottom of the screen. Following this, the
target word was presented replacing the cross and remained on the screen until
response. The responding scale was kept in the screen for the full duration of each
trial. Stimuli were presented on a 14 monitor connected to a Pentium computer.
Presentation of stimuli as well as data recording was carried out using MEL software
(Schneider, 1988). A total of 432 trials were run in three blocks of 144 trials each (72
congruent and 72 incongruent).
Experiment 1b. Control task
In the baseline black version of the task, the same number of trials was used
but this time all the words were presented in black color against a grey background.
The procedure used for this experiment was the same as in the previous study. Only
MA ran this experiment. Since there was double the amount of trials per experimental
condition, the experiment was run only once.
In order to be able to analyze together the data from MA and the control group
for the evaluation task, we used items (words), instead of participants, as the random
factor in all the analyses. Since MA run the Evaluation task twice in order to have a
more consistent estimate, her data from both sessions was averaged after we had
checked that the pattern of results was the same for both measurements. The data
from the control group was also averaged across participants to arrive at a single score
for each item presented.
A mixed 2 x 4 ANOVA was carried out with Group (MA. vs. Control participants)
and Congruency (congruent vs. incongruent) as within item factors and Valence
128
(positive, neutral, anxiety related negative and anger related negative words) as
between item factor. First, we found that our groups were not different in the overall
mean rating of words (F
(1, 68)
=1.29, p=.261). As expected, the effect of word valence
was significant (F
(3, 68)
=109.42, p=.001) and the common pattern of ratings for positive,
negative and neutral words was found.
However, the most important result was that, as predicted, color congruency
modulated emotional rating in MA but not in the control group (F
(1, 68)
=90.27, p=.001).
The control group rated words similarly independent of their color congruency (mean
rating for congruent=-0.61 vs. mean rating for incongruent=-0.59; F<1)
1
. However, as
can be seen in Graph 1, MAs emotional ratings were highly influenced by the
congruency between the color and the photism (congruent=0.22 vs. incongruent=-1.68,
F(1, 68)=90.27, p=.001). Furthermore, they were significantly more positive than the
Control Groups for the congruent color condition (F(1, 68)=17.95, p=.001), and more
negative for the incongruent color condition (F(1, 68)=184.70, p=.001).
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Control s M.A.
V
a
l
e
n
c
e

j
u
d
g
e
m
e
n
t
e
s
Congruent Incongruent Black

Graph 1. Experiment 1a and 1b. Congruency effect as a function of group. The black bar
shows the results of Experiment 1b where all words were presented in black. Error bars
denote standard error of the mean.
Interestingly, however, as can be seen in Graph 2, this congruency effect was
independent of valence (F<1) Thus, MA seemed to use two sources of information in

1
Considering the criteria used to select the words was that the average rating would be 2 for positive words, 0 for
neutral and -2 for negative words, the expected mean rating would be -0.5 since there was double the amount of
negative words than neutral or positive words
129
order to determine word valence: word meaning and congruency between the
presented color and the experienced photism. Needless to say, the same was clearly
not true for the control participants.
This congruency effect, which was only observed in MAs ratings, led to other
effects. Incongruent color produced a reduced rating for positive words and a more
accurate rating for negative words whereas congruent color resulted in an increased
rating for negative words and a more accurate rating for positive words. This led to an
overall less extreme rating in MA than the control group (F
(3, 68)
=31.66, p=.001) (see
Graph 2). It is important to note though, that the main effect of word valence was still
significant both for the control group as well as for MA (F
(3, 68)
=254.86, p=.001 and F
(3,
68)
=13.31, p=.001 respectively). Therefore, we can conclude that MA, as well as the
control participants, were following the instructions that emphasized ignoring the color
and concentrating only on the word meaning.
-3
-2
-1
0
1
2
3
Neut r al Pos it ive Negat ive
Anxiet y
Negat ive
Anger
Anxiet y
Negat ive
Anger
Cont r ols M.A.
V
a
l
e
n
c
e

J
u
d
g
e
m
e
n
t
s

Graph 2. Experiment 1a and 1b. Valence ratings. Mean valence ratings for control group,
and MA Error bars represent the standard error of the mean.
Even though the main effect of valence showed that MAs judgments were fitted,
in a control experiment we checked that her valence ratings would not differ from those
of the control participants when color-photism congruency was controlled. Here we
asked her to rate the same set of words but now all of them were presented in black
ink. Previous experiments carried out in our lab had shown that in a blocked
presentation of black graphemes or words, black was interpreted by MA as a neutral
color (i.e. did not produce an affective reaction) whereas the same black stimuli, when
130
mixed within a block of trials with colored words, were interpreted as an incongruent
color and thus produced the same pattern of results than incongruent color words.
As expected, now that only one source of information was present (i.e. word
meaning), MAs valence ratings did not differ from those of the control participants (F
(2,
136)
=2.72, p=.069). MA rated black stimuli according to their semantic valence, in a
comparable way as control participants rated both congruently and incongruently
colored words. As shown in Graph 1, now her mean rating for black stimuli was
statistically different from her own rating under color-photism congruency conditions
(F
(2, 136)
= 65.10, p=.001). Planned comparisons showed that MAs rating for black
stimuli was lower than for congruently colored stimuli (F
(1, 68)
= 14.19, p=.005) and
higher than for incongruently colored stimuli (F
(1, 68)
=220.24, p=.001).
The pattern of results observed in these experiments suggests that, in contrast
to control participants, MA seems to use two sources of information to judge
emotionality of the words: semantics and synesthetic congruency. The absence of
interaction between these two emotionality sources (lack of word valence x color
congruency interaction) suggests an independent activation of them. However, if the
photism automatically activated an affective reaction, then we would expect it to
compete with the word meaning in those conditions where both valences are
inconsistent (i.e. congruently colored negative words or incongruently colored positive
words). This would in turn raise a competition that would produce an interference
effect. To test this hypothesis, we designed a second experiment.
EXPERIMENTS 2a AND 2b. CATEGORIZATION TASK
In these new experiments we tried to reproduce the conditions that would lead
to a competition between word meaning and color congruency if the affective reactions
associated to inconsistent stimuli are automatically evoked when processing a word.
We designed a task that was a modification of the Stroop paradigm (Stroop, 1935). We
asked our participants to perform a speeded RT valence-categorization task to test our
prediction that if the postulated affective reaction is automatic, under time pressure
conditions we should find an interaction between congruency and semantic valence. In
a regular Stroop task, the irrelevant dimension of the stimulus is the word meaning
whereas the relevant dimension is the color in which the word is presented. Here, the
irrelevant dimension was the color in which the word was presented and ultimately the
congruency between that color and the photism for that word, which would lead to a
131
positive or negative affective reaction. The relevant dimension was the semantic
(emotional) meaning of the word.
In the Stroop literature, the usual finding is to observe a high level of
interference when participants are to name the ink color (less practiced task) while
avoiding to name the word (a highly practiced task). Therefore we should only expect
to find an interference of congruency on MAs valence categorization if processing the
emotion elicited by color-photism consistency was as automatic as processing of the
semantic meaning of the word.
METHOD
Participants
Again, MA and a different set of 11 non-synesthete undergraduate students run
this experiment. They were again matched for age and gender to our synesthete
participant.
Experiment 2a. Categorization task.
Only the 18 positive and 18 anger-related negative words were used in this
experiment. Again all words were presented both in congruent and incongruent color.
Twelve blocks of 36 trials were run.
Participants had to categorize the words as being either positive or negative
according to their meaning. A speeded response was required by pressing as fast as
possible one of two buttons depending on the valence of the word (i.e. positive or
negative).
Again, a white fixation point was presented against a gray background and
followed by the target word. The word was presented until response was given. Both
speed and accuracy were emphasized. Furthermore, a feedback (a tone) was
delivered for incorrect responses in order to ensure that the categorization was done
according to the meaning of the word, and not depending on the color in which it was
displayed. Again, MA run the experiment twice while control participants did it only
once.
132
Thus, the design of this experiment was a 2 (Group; MA vs. Control Group) x 2
(Color-Photism Congruency; congruent vs. incongruent) x 2 (Valence; positive vs.
negative). The first two variables were manipulated within items, whereas the last one
was manipulated between items.
Experiment 2b. Control task.
For the baseline black version of the task, everything was the same as in
Experiment 2a except that now all the words were presented always in black color
against a grey background. Again only MA ran this experiment, and only once.
Reaction Time Analysis
Before analyzing the data and once we had checked that MAs pattern was the
same in both sessions, we averaged her reaction times across sessions as well as
those of the control group across participants. A mixed Group x Valence x Congruency
ANOVA was performed on the data using items as the random factor, being the first two
variables manipulated within items and the last one between items. Trials with an
incorrect response and those with a RT above 2000ms were excluded from the
analysis (7.61% and 0.74% respectively). The following results were obtained. A main
effect of Group (F
(1, 34)
=11.70, p=.002) informed that MA was around 50ms slower than
our control group. This could be due to the fact that, under time pressure conditions,
this categorization task was more difficult for her than it was for the control group since
she had to ignore, not only the color of the words as control participants had to, but
also the correspondence between this color and the photism that has been previously
found to be automatically activated by the grapheme (Dixon et al., 2002; Elias, Saucier,
Hardie & Sarty, 2003; Mattingley et al., 2001). The participants subjective reports also
supported this idea.
Importantly, the predicted second order interaction between Group, Valence and
Congruency was indeed significant (F
(1, 34)
=6.49, p=.016) showing that for control
participants, Congruency did not affect Valence (F
(1, 34)
=1.03, p=.317). However, it did
so for MA (F
(1, 34)
=5.51, p=.025). As can be observed in Graph 3a, planned
comparisons showed that the time taken to categorize positive words was much larger
133
when they were incongruently colored than when they were congruently colored (F
(1,
34)
=6.59, p=.015). Although the pattern of results for negative words was the opposite
one, the difference between congruent and incongruent conditions was not reliable
(F<1).
We also analyzed the data from the control experiment with black stimuli and
found that MAs performance was similar to that of control participants (F<1), whereas it
interacted with her own data under color conditions (F
(2, 68)
=3.70, p=.030).
600
650
700
750
800
850
900
Pos it ive Negat ive Pos it ive Negat ive
Cont r ols M.A.
C
a
t
e
g
o
r
i
z
a
t
i
o
n

t
i
m
e

(
m
s
)

0%
10%
20%
30%
40%
50%
Positive Negative Positive Negative
Controls M.A.
E
r
r
o
r

R
a
t
e
s

(
%
)

Graph 3. Experiment 2a and 2b. A). Mean reaction time data. B). Mean error rate data.
Error bars represent the standard error of the mean.
In order to further analyze the data, we coded a new variable called
Correspondence. The crossing of these two variables (i.e., semantic Valence and
Color-Photism Congruency), provided us with this third variable. Corresponding trials
were those where the evaluation of both variables went in the same direction (i.e.
positive-congruently colored words and negative-incongruently colored words). Non-
corresponding trials were those where the evaluation of both variables yielded opposite
134
results (i.e. positive-incongruently colored words and negative-congruently colored
words).
We performed a new ANOVA with Correspondence and Group as independent
variables and found that there was indeed a significant interaction between both
variables (F
(1, 35)
=6.55, p=.015). For the control group, the difference between
consistent and inconsistent words was not significant (F
(1, 35)
=1, p=.324) whereas for
MA we found that inconsistent words were more difficult to categorize, as shown by
much longer reaction times (mean RT=787ms), than consistent words (mean
RT=729ms), F
(1, 35)
=5.41, p=.026. Therefore, it seems as though color Congruency
does have an effect in valence categorization and more specifically, the
Correspondence between color Congruency and semantic Valence seems to explain
the different pattern of results found in MA and the control participants. Therefore, an
incongruently colored word could be easier to categorize if the connotative meaning
points in the same direction as the implicit evaluation carried out over the color features
of the word. Converging evidence for this idea comes from the analysis of errors.
Accuracy Analysis
The same ANOVA was carried out for the percentage of errors. There was again
a main effect of group (F
(1, 34)
=49.72, p=.001). MA was more error prone than our group
of control students. Together with the RT data this suggests that in fact the task was
more difficult for her than it was for the other participants, as she had to solve the
incongruence introduced by the coloring of the words.
Although we found that there was no main effect of Valence (F
(1, 34)
=2.33,
p=.136) there was an interaction by group (F
(1, 34)
=6.78, p=.014) in the sense that MA
had more trouble categorizing positive words than negative words.
There was a main effect of Congruency (F
(1, 34)
=6.78, p=.014). Congruently
colored words were more correctly categorized than incongruently colored words and
this was so for MA as well as controls. Nevertheless, the most interesting result was
the significant interaction between Valence and Congruency (F
(1, 34)
=19.00, p=.001).
Incongruently colored positive words seemed to be more error-prone than any other
type of word. Adding to this, and more importantly, the predicted second order Group x
Valence x Congruency interaction was also significant (F
(1, 34)
=32.15, p=.001). We then
proceeded to separately analyze each group and found that the described pattern of
results was only found in MA (F
(1, 34)
=26.08, p=.001). Again, more categorization errors
135
were made for positive words when they were incongruently colored (F
(1, 34)
=27.03,
p=.001) and the opposite pattern was found for negative words. Fewer categorization
errors were made for incongruently colored negative words than for congruently
colored ones (F
(1, 34)
=4.01, p=.052).
Even though there was also a significant interaction between Congruency and
Valence for the control participants (F
(1, 34)
=12.40, p=.001), it was just due to the fact
that negative words yielded more errors when incongruently colored than when
congruently colored (p<.001). As can be seen in Graph 3b, it is worth noting that this
pattern of results is the exact opposite to that found in MA where negative words were
more difficult to categorize when congruently colored. Moreover, the magnitude of the
difference in the control group was much smaller. Although no clear explanation is
available for this significant interaction, it could be due to general color associations in
non synesthetes.
Regarding the control experiment in which only black colored words were
presented to MA (Experiment 2b) we found that her performance was statistically
different from that obtained when she responded to congruently and incongruently
colored words (F
(2, 68)
=21.23, p<.001) and not different from the control group (F<1), so
again we see how the additional information coming from the color Congruency can
account for the differences seen between MA and the control participants.
Finally we re-coded the variables and analyzed the percentage of errors as a
function of Correspondence and Group. Once again, there was an interaction between
Correspondence and Group (F
(1, 35)
=30.02, p=.001). A deeper analysis showed a large
correspondence effect for MA in the same direction as that found for RT: non-
corresponding words were more difficult to categorize (mean error=29%) than
corresponding ones (mean error=13%) (F
(1, 35)
=23.27, p=.001). For control participants
we also found an effect, although again it was in the opposite direction and of a much
smaller magnitude (i.e. more errors for corresponding words than non-corresponding
words (F
(1, 35)
=11.03, p=.002) (see Graph 3b).
The results from these new experiments point in the direction of an automatic
activation of the synesthetic valence of words. As can be seen in Graph 3a and 3b,
MAs reaction time and error percentage raise considerably in the non-corresponding
conditions, especially in the positive-incongruent one. Given that feedback was
provided after each trial, this pattern of data suggests that MA was making a
considerable control effort to respond correctly, but the synesthetic valence (particularly
136
the incongruence) exerted an influence to the extent that the correct response could
not be controlled at the proper levels. In a standard Stroop experiment, word meaning
is taken as the automatic dimension and color naming as the intentional one. The
subject has to name the color and inhibit the response associated to the potent
activation that comes from the automatic processing of the denotative meaning of the
word. Here, the intentional response is the connotative meaning of the word, a
dimension that uncountable experiments have demonstrated to be automatically
processed (Catena, Fuentes & Tudela, 2002). Moreover, the criterion to select the
words was based on previous experiments to ensure they were clearly positive or
negative. Still, the affect produced by the incongruence of the color in which the word is
presented overcomes that activation of the word meaning and influences the final
response. Previous studies carried out in our lab have shown that this interference from
the affective component of a word in a categorization task is unusual even when the
subjects are trait-anxious patients.
GENERAL DISCUSSION
In these series of experiments we have shown for the first time to our
knowledge, that a) emotion is part of the set of subjective experiences known as
synesthesia and b) its effects on behavior may be as clear and robust as those
produced by the concurrent (i.e. the photism) itself.
In Experiment 1 we showed that color Congruency does influence emotionality
ratings in parallel with word Valence. MAs ratings were accurate in the sense that she
rated words according to their meaning as control participants did. But she was also
influenced by the color in which the word was presented since congruently colored
words were rated more positive than incongruently colored words. If color Congruency
does have an effect on word Valence and it is of an automatic nature, we would predict
that it would affect performance under conditions where a fast response is requested
and no time is available for control mechanisms to completely inhibit the information
coming from this irrelevant dimension of the stimulus.
Experiment 2 was designed to test this hypothesis and we found that in fact,
when time pressure was introduced, Color-Photism Congruency affected performance
in a very interesting way. By presenting words with semantic valence colored either
congruently or incongruently with the synesthetes set of photisms, we have shown the
presence of a bias causing participants to respond faster when there is a consistency
137
between the 2 dimensions of the presented stimulus (i.e. Valence and Congruence).
That is, synesthetes might be faster responding to an incongruently colored word than
to a congruently colored one depending on the semantic meaning of it.
Previous studies have shown that the perception of the photism associated to a
particular word is highly automatic (Dixon et al., 2000; Lupiez & Callejas, 2006).
Affective priming studies have also shown participants ability to automatically extract
connotative meaning of words even when they do not have to be consciously evaluated
(Fazio, Sanbonmatsu, Powell & Kardes, 1986). Putting this together, only if the
affective reaction associated to the match-mismatch between a words photism and the
presented color happens automatically would one expect the affective reaction to
interact with the automatic evaluation of the words connotation. And only if that
affective reaction was strong enough, would it influence such an automatic processing
of words. Moreover, only if this affective reaction is sufficiently strong, would the control
mechanisms fail to suppress it in order to give an accurate categorization of the words
based on their connotative meaning. It seems then that assessing the fit between the
environment and the subjective experiences known as synesthesia, as well as the
affective reaction associated with such evaluation, is a strong and highly automatic side
effect of at least grapheme-color synesthesia
1
. Also, in a study carried out in our lab in
which the task at hand was an abstract/concrete categorization of these same words,
that is, a task that is orthogonal to the affective processing of the words meaning, we
found the same pattern of results where incongruently negative words and congruently
colored positive words are categorized faster than congruent-negative and
incongruent-positive ones. It seems then as if synaesthetes automatically evaluate
whether the external physical words fit their internal coloring schema. This evaluation
seems to happen even when it is not relevant to the task at hand.
From this perspective, the next step in the study of emotions as a side effect of
synesthesia would be to discern the mechanisms underlying such processes and
whether they are shared by non-synesthetes. Previous studies have pointed to the
restrosplenial cortex as a structure that could be interesting to study (Ward, 2004). This
structure has been related to the processing of familiar people as opposed to unfamiliar
people and also to the processing of emotional words as compared to neutral words

1
Although not reported here, several other synesthetes have been tested and show the same pattern of response
times for the categorization task.
138
(Maddock, 1999 for a review). Also, this area appears to be active in different studies of
brain activation and synesthesia (Nunn et al., 1999, Weiss et al., 2001). Although this
area has also been related to memory, a recent study showed that, controlled for
memory variables, this area was found active when categorizing words according to
their valence (Maddock, Garrett & Buonocore, 2003). Knowing when these evaluative
processes take place could be taken as an index of the processing stage where
subjective colors (photisms) are bound to the percept and consciously experienced.
As previously mentioned, information that arrives to V1 by means of the retina is
then relayed to V2, V4 and temporal areas and it is then that it travels to the amygdala
(Amaral et al., 1992). Posterior cingulate cortex-retrosplenial cortex does not have
direct connections to the amygdala but it does connect to other frontal regions also
related to emotional evaluation such as the anterior cingulate cortex. It also receives
input from the orbital and dorsolateral prefrontal cortex, parahippocampal cortex,
superior tempral sulcus, precuneus, claustrum and anterior and lateral thalamic nuclei
(Goldman-Rakic, Selemon & Schwartz, 1984; Van Hoesen, Morecraft & Vogt, 1993).
The question then would be whether higher synesthetes, whose proposed locus for
cross-activation is higher up in the visual processing, next to the angular gyrus, would
show the same pattern of responses as lower synesthetes whose proposed locus of
cross-activation would be localized around hV4 (Ramachandran & Hubbard, 2001b).
A recent study points to the possibility that some linguistic features of words,
such as their emotional valence, might be processed very early in time (Ortigue et al.,
2004). These authors found ERP differences at around 150ms for emotional words
compared to neutral words when the task at hand was discriminating words from non-
words. Other electrophysiological studies suggest that the information about the
meaning of the word can be accessed almost as the same time as information about its
form (Pulvermller, 2001). It could be the case that emotional words also had a special
representation and could be accessed faster than equal frequency neutral words. If so,
we could then predict that the effect of incongruently colored words would be present
only in those cases where photisms were bound to graphemes prior to graphemes
emotional assessment (i.e. only for projector-lower synesthetes; Dixon, Smilek &
Merikle, 2004; Ramachandran & Hubbard, 2001b). If the information is sent before the
photism is added to the phenomenological experience or if no photism information is
sent (i.e. as in the case of non-synesthetes) the outcome of the evaluation would be
the usual one. However, if words meaning is not accessed so early in the processing, it
might be the case that by the time words information is relayed to the emotional
139
centers for its evaluation, it has already been dressed with its corresponding photism
by all synesthetes independent of the stage at which the cross-activation is occurring. If
the object of evaluation includes the information about the photism, then the evaluative
process will have to deal with the semantics of the word itself, the inconsistencies
between the 2 pieces of color information and the relationship between these two
features of the stimulus being assessed. We believe that MA, as well as the other
synesthetes tested, is a higher-associator synesthete. However, informal observations
point to the possibility that lower-projector synesthetes also show emotional reactions
associated to the perception of incongruently colored stimuli. More investigation needs
to be carried out on the dynamics of this affective reaction to clarify the specific
characteristics of the synesthetes showing it and the temporal course of the stimulus
processing when congruent and incongruently colored.

CHAPTER 4
ARE ALL SYNESTHETES CREATED EQUAL ?
GENERALIZATIONS OF SYNESTHETICALLY
INDUCED AFFECTIVE REACTIONS

Chapter 4. Generalization Of Affective Reactions In Synesthetes _____________________
143
ABSTRACT
We have previously found that the perception of an incongruently colored
stimulus such as a word gives rise to an affective reaction that in turn influences
emotional rating performance in our synesthete MA. Here we try to replicate those
findings in a sample of three other synesthetes to check whether the results previously
found were due to idiosyncrasies of our synesthete. A categorization task where
emotional words were presented in either the congruent color or an incongruent one
was administered to our three synesthetes. Results show that the previously found
pattern was replicated in tow of the three synesthetes. A subsequent set of experiments
were performed to check whether the synesthesia in the synesthetes not showing the
effect was too weak as to affect performance. Results informed that strength of
synesthesia for word and some personality traits could be driving the absence of
congruency effect.
INTRODUCTION
It now seems to be an accepted statement that not all synesthetes are equal
and not all of them experience their synesthesia in the same manner (Dixon, Smilek &
Merikle, 2004; Dixon & Smilek, 2005; Hubbard, Arman, Ramachandran & Boynton,
2005; Hubbard & Ramachandran, 2005; Ramachandran and Hubbard, 2001b; Smilek
& Dixon, 2002). However, there are still some researchers that do not acknowledge this
difference or argument that our knowledge about synesthesia is still far away from a
time when clear factors affecting one synesthetes and not other can be identified and
tested (Edquist, Rich, Brinkman & Mattingley, 2006; Mattingley, Payne & Rich, 2006).
Dixon et al. (2004) tested a group of synesthetes on a Stroop color naming task
and a Stroop photism naming task and found that the pattern of results for what they
call projectors was reliably different from the pattern found for the associators. Most
important, the groups had been defined a priori based on the subjective reports of the
synesthetes tested about the peculiarities of their synesthetic experiences. Also,
Hubbard et al. (2005) found that the extent to which hV4 was activated in synesthetes
correlated with their performance in behavioral tasks used to evaluate synesthesia.
They also found that 2 distinct groups could be differentiated based upon both, the
behavioral performance and the brain activation level.
__________________________________ SECTION 2. SYNESTHESIA AND EMOTIONS
144
Nevertheless, the specific determinants of the different types of synesthesia are
yet to be enunciated. Although some synesthetes readily volunteer that they
experience colors out there in space or in their minds eye, there are still some other
synesthetes that do not fit into these categories and explain their synesthesia as
neither of the previous two. Whether this means that the higher/associator-
lower/projector classification is not complete or that it is focusing on the wrong features
of the experienced phenomena is yet to be discovered. What seems to be clear is that
independently of the features that describe each subgroup of grapheme-color
synesthesia, synesthetes are a heterogeneous group and that fact has to be taken into
account in order to make scientific progress.
A quite common but virtually unstudied feature of grapheme-color synesthesia
is the emotionality attached to the synesthesia experience itself. When a stimulus color
matches the photism that a synesthete experiences for it, a positive sensation is felt.
On the other hand, if the color does not match the internal experience a discomforting
sensation is felt. This side-effect of grapheme color synesthesia was first empirically
reported by Callejas, Acosta & Lupiez (submitted, Chapter 3). A congruency effect
was found so that congruently colored words (i.e. same color as the synesthetes
reported photism for the word) were categorized faster than incongruently colored
words. However, this was only found for positive words and an absence of congruency
effect or a reversal was found when the stimulus eliciting a color was a negative
valence word. The authors proposed that synaesthetes automatically evaluate whether
the external physical words fit their internal coloring schema and this would happen
even when it is not relevant to the task at hand. Therefore external coloring of the
words being either congruent or incongruent with their internal coloring schema led to
positive or negative affective reactions, which interfered with the evaluation of the
emotional meaning of the words.
The study presented here had two goals. On the one hand we wanted to test
whether the affective reactions found in our synesthete MA when viewing stimuli
colored incongruently with her photisms (Callejas et al., submitted, Chapter 3) are
common to other synesthetes. On the other hand we wanted to check whether
differences would be found amongst synesthetes in this respect.
The fact that a wide number of synesthetes that communicated with us as well
as with other labs have reported feeling discomfort when watching a stimulus wrongly
colored gave us a relative certitude that the results previously found would be
replicated with other synesthetes. However, it could also be the case that some
145
idiosyncratic features of MAs perception or personality were making that discomfort to
be reflected in her behavioral performance. Therefore we conducted the present set of
experiments to test these synesthetically elicited affective reactions in other
synesthetes different from MA.
EXPERIMENT 1a AND 1b. EFFECT OF PHOTISM-COLOR CONGRUENCY IN
A CATEGORIZATION TASK.
We first tried to replicate the findings of Callejas et al.s Experiment 2a and 2b
on a set of 3 different synesthetes. In the original study, a set of positive and negative
emotionally charged words were presented in congruent and incongruent color
(Experiment 2a) or in black color (Experiment 2b) and the participants had to
categorize them as positive or negative according to their semantic connotation, while
ignoring the color itself. What they found was that Photism-Color Congruency
interacted with word Valence so that corresponding conditions (positive-congruently
colored and negative-incongruently colored) were faster and more accurately
responded to than non-corresponding conditions (positive-incongruently colored and
negative-congruently colored). When words where presented in black, however, the
pattern of results mirrored that of control participants. We run these same experiments
in three other synesthetes that also experienced colors for words and we compared
their subjective reports with the objective performance.
METHOD
Participants
Three synesthetes and three groups of control participants performed these two
experiments.
EL is a grapheme-color synesthete that was 16 years of age at the time of
testing. She consistently perceives colors for letters, numbers and words as a whole.
She also experiences other types of synesthesia. She experiences colors in her minds
eye. Prior to testing she reported not caring much whether the words were correctly or
incorrectly colored.
PR is a grapheme-color synesthete that was 52 at the time of testing. As the
previous synesthetes, she also experiences color in her minds eye for letters, numbers
and words as a whole. She reported disliking incongruently colored words.
146
Both EL and PR are Canadian and native English speakers. They were
contacted by means of the Waterloo Synesthesia Research Centres database and
tested at the University of Waterloo, Canada.
PSV is a native Spanish synesthete. She is 29 and among other types of
synesthesia, she experiences colors for numbers, some letters and words, in her
minds eye. She is a Fine Arts graduate student and as such, she has a very fine
grained ability to discriminate and classify colors. She reports a strong displeasure
when seeing something incorrectly colored. She explains it as something similar to
the sensation felt when watching something disgusting that automatically drives oneself
to fixate at a different object or location.
The control participants for each group were native speakers of the same
language as the synesthete they were paired with. Six English speaking participants
run ELs and PRs experiments. Nine Spanish speaking controls run PSVs
experiments.
A set of 18 positive words and 18 negative words were selected for EL and PR
from the ANEW (Bradley & Lang, 1999) taking into account Word Frequency, mean
Valence and strength of the photism experienced for them. Both Frequency and
Valence values were drawn from the ANEW. Word mean frequency was 44 (SD 50).
Frequency was not significantly different between EL and PR, neither between positive
and negative words (both Fs<1). Word mean valence for positive words was 8.2 for EL
as well as PR. For negative words it was 1.8 again for both of them (ANEW min: 1.25;
max: 8.82). Therefore words were clearly positive or clearly negative. Last, the
subjective strength of photism perception measured in a 7 point scale (1: very faint; 7:
very vivid) was, for EL: 7 for positive words and 6.3 for negative words, and for PR: 6.5
for positive words and 5 for negative words.
For PSV out of the 72 words used in the Callejas et al. study, a set of 26 were
selected so that neither of them elicited a black or a grey photism. A total of 13 positive
words and 13 negative words were selected. Words were equated in subjective
familiarity (4.6 in both cases), with a mean emotional rating of -3.8 for negative words
and 3.5 for positive words as measured by previous studies in control participants in a
10 point Likert scale (from -5 to +5). PSV also rated the vividness of her photisms in a
7 point scale. For positive words it was 5 and for negative words it was 6.4.
147
Words were presented on a neutral grey background screen to ensure good
visibility of all different colors.
Programming of the experiments, stimulus presentation and data collection
were carried out on a 14 screen controlled by a Pentium computer running e-prime
(Schneider, Eschman & Zuccolotto, 2002)
Four variables were present in Experiment 1a. Valence had two levels (positive
and negative words) and Congruence had two levels (congruent and incongruent
color). A variable Group was coded with two levels: synesthete and control. Last, since
three different synesthetes were tested along with their corresponding control group, a
new variable termed Case was coded with three levels: EL, PR and PSV or ELs, PRs
and PSVs control group.
Congruent color was the color chosen by each synesthete for each word.
Incongruent color was the color-wheel opposite to the congruent one. For each
synesthete and its corresponding control group a total of 8 trials per word and coloring
condition were presented. Therefore EL and PR run 576 trials each (4 blocks of 144
trials each) and PSV run 416 (8 blocks of 56 trials each). Before the experimental trials
a set of practice trials were given to familiarize synesthetes and controls with the task.
Participants were encouraged to rest between blocks.
In Experiment 1b all words were presented in black ink and therefore only
Valence was manipulated. Only synesthetes run this experiment. Again 8 trials per
word were presented (4 blocks of 72 trials each for EL and PR and 4 blocks of 52 trials
for PSV).
Each trial began with a fixation point, which was presented for 1000ms and
followed by the target word, which replaced the fixation point. Participants had to hit
one of two keys (i.e. z and m) depending on whether the word was positive or
negative. Auditory feedback was given when an incorrect response was emitted. All
synesthete participants run Experiment 1a first and Experiment 1b after a short break.
Trials with incorrect responses were eliminated from all the RT analyses. In
Experiment 1a these accounted for 5.3% and 4.2% (EL and her control group
148
respectively); 1.7% and 1.7% (PR and her control group respectively) and 1.7% and
6.6% (PSV and her control group respectively). For Experiment 1b 1.7% was excluded
for EL, 2% for PR and 0.5% for PSV. Since participants reaction times were somewhat
variable, a 2 standard deviation cut off criteria was set to eliminate extreme values.
Items (words) and not participants were used as random factor in the statistical
analysis. Therefore, for each word and group, reaction times and error percentage
were averaged -across participants and trials- for each control group and -across trials
only- for each synesthete for each experimental condition (word valence and
congruence type). Since the words used for each synesthete and control group were
different, Case along with Valence was analyzed as a between items variable. Group
and Congruence were considered within item variables.
Experiment 1a. Reaction time analysis
A mixed ANOVA was carried out with Group (synesthete and control), Case (EL,
PR and PSV or ELs, PRs and PSVs control group), Valence (positive and negative
words) and Congruence (congruently and incongruently colored words). Mean reaction
time and percentage correct per condition and group can be found in Table I.
Table I. Experiments 1a and 1b. Mean categorization RT per group, subject and condition.
Error percentages are shown in parenthesis.
Group
Synesthetes Controls
Valence Congruence EL PR PSV Total ELs PRs PSVs Total
Positive Congruent
607.2
(3.5%)
573.2
(2.1%)
640.5
(2.9%)
607.0
(2.8%)

637.67
(3.5%)
646.5
(2.4%)
658.3
(7.9%)
647.5
(5.2%)
Incongruent
607.2
(9.0%)
623.4
(2.1%)
645.0
(1.0%)
625.2
(4.0%)

641.1
(3.1%)
646.9
(2.8%)
662.1
(6.6%)
650.1
(4.7%)
Black
558.6
(2.8%)
626.4
(0.7%)
622.1
(0.0%)
602.4
(1.2%)

Negative Congruent
661.44
(5.6%)
620.90
(1.4%)
622.77
(2.9%)
635.0
(3.3%)

621.4
(3.1%)
630.5
(1.4%)
680.3
(6.4%)
644.1
(3.7%)
Incongruent
735.2
(3.5%)
616.2
(1.4%
615.1
(0.0%)
655.5
(1.6%)

630.4
(4.9%)
642.1
(3.1%)
695.0
(5.7%)
655.9
(4.6%)
Black
565.6
(2.8%)
623.8
(1.4%)
607.7
(1.0%)
599.0
(1.7%)

Positive words on average were categorized as fast as Negative words (F<1).
Synesthetes as a group where not faster than controls (F
(1, 92)
=2.09, p=.152). However,
a significant Group x Case interaction showed that this was due to EL (F
(2, 92)
=22.89,
p=.001) being slower than the other synesthetes (both ps<.05) and slower than her
149
control group (p=.015) while the other two synesthetes were as fast as their respective
control groups. Last, as expected, there was a congruency effect (F
(1, 92)
=20.36,
p=.001) modulated by Group (F
(1, 92)
=10.13, p=.002). As shown in Graph 1, the
congruency effect was found for the synesthetes (F
(1, 92)
=17.84, p=.001) but not for the
controls (F
(1, 92)
=3.12, p=.080), although they showed a trend in the same direction.
Congruently colored words were responded to faster than incongruently colored words.
550
600
650
700
750
EL' s PR' s PSV' s EL PR PSV
Controls Synesthetes
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)

Graph 1. Experiment 1a. Reaction times for each Group and Case as a function of
Congruence.
The second order interaction by Case was also reliable (F
(2, 92)
=9.08, p=.001). A
separate ANOVA for each Group showed that the three cases of control participants
performed in a similar manner (F<1). Synesthetes on the other hand showed different
susceptibility to the Congruence effect (F
(2, 92)
=10.14, p=.001). While planned
comparisons informed that EL showed faster reaction times for congruently colored
words (F
(1, 92)
=40.26, p=.001), PR showed a Congruence effect although it was not
statistically reliable (F
(1, 92)
=2.34, p=.129) and PSV did not even show nominal
differences (F<1). Last, the second order interaction between Congruence, Valence
and Group was also significant (F
(1, 92)
=5.30, p=.023, see Graph 2).
A separate ANOVA for each group showed that, as expected, Controls did not
have a significant interaction between Congruence and Valence (F(1, 92)=1.30,
p=.257) whereas Synesthetes did (F
(1, 92)
=3.91, p=.051). As shown by planned
comparisons there was a congruency effect for positive words (F
(1, 92)
=19.22, p=.001)
and no effect for negative ones (F
(1, 92)
=2.53, p=.115). Last, the Congruence x Valence
interaction was not different for the three groups of controls (F<1) nor was it for the
three synesthetes (F<1). However, a separate analysis for each synesthete (see Graph
3) showed that the interaction between Congruence and Valence was only present in
150
PR (F
(1, 34)
=10.46, p=.003) but not in EL or PSV (both Fs<1). PR showed a congruency
effect for positive words (F
(1, 34)
=17.46, p=.001) and a nominally inversed congruency
effect for negative words (although not statistically reliable, F<1). Planned comparisons
showed that, although the interaction was not reliable for EL, the magnitude of the
Congruence effect was larger for positive words than for negative words (F
(1, 34)
=13.65,
p=.001 and F
(1, 34)
=5.66, p=.023 respectively)
500
550
600
650
700
Posivite Negative Posivite Negative
Control Synesthetes
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)

Graph 2. Experiment 1a and 1b. Reaction times for each Group as a function of Valence
and Congruence.
500
550
600
650
700
750
800
Posivite Negative Posivite Negative Posivite Negative
EL PR PSV
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)

Graph 3. Experiment 1a and 1b. Reaction times for each synesthete as a function of
Valence and Congruence.
Results from a previous rating experiment (similar to Experiment 1a in Callejas
et al., submitted, Chapter 3) are consistent with this categorization data. Participants
were asked to rate how positive or negative the same set of words was on a 7 point
Likert scale. Contrary to MAs experiment, here a speeded response was required. A
Congruence effect was found for EL and PR and this effect interacted with word
Valence. Congruently colored words were given more accurate ratings than
151
incongruently colored ones and this effect was maximal for positive words. PSV
showed no difference as a function of Congruence.
Experiment 1b. Reaction time analysis
In order to check whether a blocked presentation of black words was also a
neutral condition for these three synesthetes we compared each synesthetes
performance on the black version of the task to their own performance under color
conditions and also to the performance of their respective control groups. Results on
the black version can be seen in Graph 3 and in Table I.
For EL we found that when congruent, incongruent and black were consider as
three levels of a Congruence variable, she showed an effect (F
(2, 68)
=42.68, p=.001)
consisting on slower times for incongruently colored words than congruently colored
ones (p=.001) and slower times in congruently colored words than in black colored
words (p=.001). When we compared her performance under black conditions with that
of her control group we found a significant effect (F
(2, 68)
=58.88, p=.001); she was faster
than controls (both ps<001).
For PR we found again an effect of Congruence (F
(2, 68)
=4.48, p=.015) and post
hoc comparisons showed it was due to the fact that congruently colored words tended
to be faster than incongruently colored ones (p=.65) and faster than black ones
(p=.017). Black words were not different from incongruent ones (p=.855). When black
performance was compared to the results of her control group we found no differences
(F
(2, 68)
=1.94, p=.152).
Last, for PSV there was no significant difference between congruent,
incongruent or black colored words (F
(2, 48)
=2.37, p=.104). Again, since black showed
the same RT pattern than congruent and incongruent colored words, it was also
statistically different than the performance of her control group (F
(2, 48)
=35.40, p=.001).
While controls performance for congruent and incongruent colors were not different
(p=.498), both were slower than PSVs performance for black words (both ps<001).
It is interesting to see how black words do not behave in the same manner for
all three synesthetes. While EL seems to be faster for black words than in any other
condition, PR shows the same performance for black words as for incongruently
colored words. PSV did not show a congruence effect and so the expected absence of
difference between black and colored words was found. PRs results could either mean
152
that black is not taken as a neutral shade and it thus resembles the RTs obtained for
the incongruent condition or that she does not have an effect of incongruence but a
large benefit for congruently colored positive words. Given her overall pattern of results
and that of her control group, it seems as though she is just obtaining a benefit for
congruently colored positive words as compared to any other condition.
Experiment 1a. Accuracy analysis
The same ANOVA was carried out on the error percentage for each group.
Mean error rates per group and experimental condition can be seen in Table I. No main
effect of Valence showed that positive words were as easy to categorize as negative
words (F<1). Synesthetes as a group were more accurate than control participants (F
(1,
92)
=4.70, p=.033). In fact a significant Group by Case interaction (F
(2, 92)
=9.74, p=.001)
showed that PSVs control group was less accurate than her (p=.036). Congruent
words were categorized as accurately as incongruent ones (F<1). See Graph 4.
0%
2%
4%
6%
8%
Posivite Negative Posivite Negative
Control Synesthetes
E
r
r
o
r

p
e
r
c
e
n
t
a
g
e

(
%
)

Graph 4. Experiment 1a. Error percentage for each Group as a function of Valence and
Congruence.
Only the expected Group x Congruence x Valence second order interaction was
significant (F
(1, 92)
=6.06, p=.016,). Separate ANOVAS for the synesthetes group and the
controls group showed a marginally significant interaction for the synesthetes F(1,
92)=3.51, p=.064 and no effect for the controls (F
(1,

92)
=1.37, p=.244). Synesthetes
were more accurate for incongruently colored negative words and congruently colored
positive words (see Graph 4). Last, the third order interaction approaching significance
(F
(2, 92)
=2.71, p=.070) showed that the Congruence x Valence x Group interaction
tended to be different for each Case.
153
0%
2%
4%
6%
8%
10%
12%
Posivite Negative Posivite Negative Posivite Negative
EL PR PSV
E
r
r
o
r

p
e
r
c
e
n
t
a
g
e

(
%
)

Graph 5. Experiment 1a. Percentage correct for each synesthete case as a function of
Valence and Congruence.
The separate ANOVA for each Group showed a lack of Case x Valence x
Congruence interaction for the controls group (F<1) and a trend towards an interaction
for the synesthetes group (F
(2, 92)
=2.70, p=.073). As shown in Graph 5, PR showed
nominally equal error rates for congruently and incongruently colored words. Planned
comparisons informed that PSV also showed no congruency effect for either positive or
negative words (F<1 and F
(1, 92)
=1.93, p=.167 respectively). Last, the pattern found for
EL was a replica of that found for the synesthete group as a whole. It was also the
same one found in previous experiments with MA (Callejas et al., submitted, Chapter
3). She showed improved performance for corresponding conditions (i.e. congruently
colored positive words and incongruently colored negative words) as compared to non-
corresponding conditions although only the effect for positive words was significant (F
(1,
92)
=9.93, p=.003 for positive words and F
(1, 92)
=1.40, p=.240 for negative words).
Experiment 1b. Accuracy analysis
Again synesthetes performance for black words was compared to their own
performance under color conditions and to that of their corresponding control group.
Error percentages for the black condition are also depicted in Graph 5.
EL showed again the biggest differences between Congruence conditions (F
(2,

68)
=4.68, p=.012). Black words were the most accurate ones and incongruently colored
words the more error prone and this difference was reliable (p=.010). Her performance
under black coloring conditions was not different from that of her control group (F
(2,

68)
=2.05, p=.136). PR showed no difference between Congruence conditions (F<1).
Black coloring did not influence response accuracy differently than congruent or
154
incongruent colors. When compared with her control group no differences were found
either (F<1). Last, PSV showed no difference as a function of color (F
(2, 48)
=2.11,
p=.132). When compared with her control group we found she was different (F
(2,

48)
=27.96, p=.001) because she was more accurate than they were.
Several interesting facts rise from this data. First, the synesthete group data as
a whole replicates previous findings (Callejas et al, submitted, Chapter 3). A significant
modulation of Congruence by word Valence was found for the synesthetes group and
not for controls group in RT data as well as in accuracy data. This result supports the
idea that the modulation of Congruence by Valence found in MA was not due to
idiosyncrasies but to a real phenomena that can be found in other synesthetes.
However, it is also important to note that not all synesthetes as individuals
showed that pattern. A priori we had two reports of disliking wrongly colored words and
one report of indifference. PR reported not liking wrongly colored items. PSV was much
stronger in her statement. She equated it with something so unpleasant as to drive
gaze away from it. Last, EL showed indifference about coloring conditions although it is
important to point out that after running the experiments she did report having found
wrongly colored words quite unpleasant to work with. Therefore, although all of them
informed disliking wrongly colored words in different degrees, behavioral data only
backs up the statements of EL and PR. While EL showed a modulation of congruency
for both RT and accuracy measures, PR only showed it for RT data. Surprisingly, PSV
did not even show a congruency effect in the first place. This will be discussed in more
detailed in a later section.
The fact that the modulation of Congruence by Valence was found in these two
synesthetes has an additional importance since they were native speakers of a
different language than MA and therefore the effect cannot be attributed to any specific
feature of Spanish language or the words used. It is also interesting to see that the
most consistent result is a decrease of the congruency effect for negative words but
rarely does the effect reverse (i.e. faster or more accurate reaction times for
incongruently colored negative words). This reversal was found in MA but not in these
two synesthetes. Since all these synesthetes reported perceiving colors in their minds
eye, the documented difference between projectors and associators (Dixon et al.,
2004) cannot be the factor accounting for these different patterns of results. Other yet
unidentified features of the synesthetes, their perceptions or the stimuli used may
account for these differences.
155
What seems clear is that, although the general pattern found for synesthetes as
a group does replicate previous results, not going further in the analysis would lead to
erroneous conclusions since some synesthetes showed the Valence modulation of
Congruence and some others did not even show a Congruence effect. Mattingley &
Rich (2004) also reported that the variability of their data for a large group of
synesthetes in a synesthetic Stroop task was considerably high and while some
synesthetes showed a 200ms effect, others did not show a congruency effect at all. In
fact, some participants even showed a reversal Stroop effect whereby when they had to
name the color in which a stimulus was presented and such color was different from
the photism elicited by the presented stimulus they were faster than when the color
matched the elicited photism. They propose that relative strength of the synesthetic
color and not a difference between associators and projectors might be the underlying
reason for such variability in results. PSVs subjective ratings for the vividness of her
synesthetic experience were no different from those of EL. However, these ratings only
ensure that each synesthetes experience is maximized but not that it is equated
across synesthetes. It could well be that PSVs experiences are in general fainter than
those of EL and PR.
Since we had not found a congruency effect for PSV in this valence
categorization task, we decided to carry out a standard synesthetic Stroop task and a
modified synesthetic Stroop task with her just as we did for MA (Lupiez & Callejas,
2006, Chapter 1) to test whether her lack of effect was due to a general lack of
interference between photisms and real colors.
EXPERIMENTS 2a AND 2b. TEST OF THE STRENGTH OF PSVs
SYNESTHETIC EXPERIENCES: STROOP INTERFERENCE WITH A COLOR
NAMING AND A PHOTISM NAMING TASK
We used a standard Stroop (Stroop, 1935) task in which participants had to
name the color of a letter or number presented on the screen. The identity of the letter
was manipulated so that it would elicit the same color presented or a different one. For
the modified Stroop task we showed the very same stimuli but now asked our
participant to name the photism elicited by the letter and to ignore the displayed color. If
her absence of congruency effect was due to very weak experience of her photisms,
we would expect an absence of standard Stroop effect for her because the experience
of photisms was not strong enough as to affect her color naming times and we would
also expect to find a strong congruency effect in the photism naming version since the
156
real color would interfere to a greater extent. Alternatively, if photisms were vividly
experienced, we would expect to find a large Stroop interference effect in the color
naming task and a smaller interference effect in the photism naming task.
In this task we also measured negative priming as the influence of a previously
ignored color on the current response to test whether PSV was able to inhibit
information that would interfere with the task being performed.
METHOD
Participants
PSV and the same set of Psychology Undergraduate Students that served as
control group for the valence categorization task performed the experiments. Obviously,
control participants only performed the color naming task (Experiment 2a) while PSV
performed both color naming and photism naming tasks.
A set of eight letters and numbers were selected so that they elicited one of four
different colors. Colors chosen were black, white, red and yellow. Each stimulus
appeared in either a congruent or incongruent color. Congruent colors were those
picked by PSV for each stimulus. Incongruent colors were the other three remaining
colors. The target stimulus subtended .76 vertical and .54-.86 horizontal.
(Schneider et al., 2002). This was the only difference with Lupiez & Callejas, 2006,
Chapter 1) since they programmed the task using MEL (Schneider, 1988). Naming
times were collected by means of a voice key and response accuracy was coded with
the keyboard.
The design of the experiment was the same as in Lupiez & Callejas (2006,
Chapter 1). Since each letter appeared in each one of the four colors, there were 25%
congruent trials and 75% incongruent trials.
157
In each trial a black fixation point appeared on a dark grey screen for 500ms
and was followed by the target stimulus. The target was present until a vocal response
was given. After accuracy was manually coded the next trial began.
Five blocks of 48 trials were run for each experiment (a total of 240 trials). There
were 60 congruent trials and 180 incongruent trials. As in Lupiaez & Callejas, (2006,
Chapter 1), since the order of the trials was pseudo-randomly assigned so that neither
the same color nor the same elicited photism would appear in two consecutive trials,
the number of control and ignored repetition trials varied for each participant.
PSV run each experiment in three different times in different testing sessions
(counterbalanced) while control participants run Experiment 2a only once.
Reaction time analysis. Experiment 2a: Color naming.
Color naming times for each condition and for each item were averaged across
participants for the control group and across sessions for PSV. Therefore items and not
subjects were used as random factor.
The first trial of each block was eliminated from analysis. Incorrect responses as
well as trials were the voice key was triggered by stimulus other than the vocal
response were also eliminated (3.8% and 1.4% respectively for PSV and 2.9% and
1.1% respectively for the control group). Responses faster than 200 ms and slower
than 1500 were also eliminated (1% for PSV and 2% for controls). Mean naming times
and percentage correct for each condition can be seen in Table II.
Stroop analysis.
A mixed ANOVA was performed on the mean reaction times with the factors
Group (PSV vs. Controls), N-1 Congruence (N-1 congruent vs N-1 incongruent) and N
Congruence (N congruent vs. N incongruent). A main effect of group showed that PSV
was slower to respond than her control participants (F
(1, 7)
=112.57, p=.001). N-1
Congruence did not affect RTs (F<1) but N-Congruence did (F
(1, 7)
=30.71, p=.001).
Congruent trials were responded to faster than incongruent ones. This pattern was due
to PSVs performance (significant N-Congruence by Group interaction, F
(1, 7)
=18.50,
p=.004). She showed a large congruency effect (113ms, F
(1, 7)
=42.04, p=.001) while
control participants did not (F
(1, 7)
=1.39, p=.277). The second order interaction was
158
marginally significant (F
(1, 7)
=4.55, p=.070). Inspection of the data showed that the PSV
showed a modulation of N-Congruency by N-1 Congruency while participants did not.
This was not surprising given that control participants did not show an N Congruence
effect to start with.
Table II. Experiment 2a (color naming) and 2b (photism naming). Mean naming times per
condition for each group and experiment. Error rates are shown in parenthesis.
N-1 Congruent N-1 Incongruent
N-Congruent N-Incongruent N-Congruent N-Incongruent N-Incongruent
Group Control Control Control Control
Ignored
repetition
PSV
Color-naming

553.78
(1.4%)
694.39
(2.5%)

579.84
(1.5%)
663.66
(7.8%)
657.29
(6.2%)
PSV
Photism-naming

548.69
(1.8%)
600.24
(6.5%)

566.97
(0.0%)
594.37
(7.0%)
602.72
(2.2%)
Control group
Color-naming

556.68
(3.8%)
570.10
(4.4%)

544.93
(2.6%)
566.27
(5.1%)
557.18
(3.8%)

A separate ANOVA for each group confirmed that for controls there was no
effect of N-Congruence or interaction with N-1 Congruence (F<1). The same analysis
showed for PSV that her N-Congruence effect was modulated by N-1 Congruence (F
(1,
7)
=15.27, p=.001). As can be seen in Graph 6, her congruency effect was larger when
the previous trial was congruent than when the previous trial was incongruent (141ms
vs. 84ms).
450
550
650
750
N-1
Congr uent
N-1
Incongr uent
N-1
Congr uent
N-1
Incongr uent
N-1
Congr uent
N-1
Incongr uent
Color Naming Color Naming Phot is m Naming
Cont r ol Gr oup PSV
N
a
m
i
n
g

t
i
m
e

(
m
s
)

Graph 6. Experiment 2a and 2b. Stroop effect. Reaction times for PSV for the Color
Naming and Photism Naming version of the Stroop task and reaction times for her
Control group for the Color Naming version.
159
These results are clear evidence that PSV does show a congruency effect in
the standard synesthetic Stroop task and this effect is quite large considering she
reports experiencing colors in her minds eye. Her performance suffers from a clear
interference when the color she perceives does not match the internally experienced
one. It is also interesting to see that her online control mechanisms, although able to
reduce the interference, are not capable to eliminate it.
Negative Priming analysis.
To further test her ability to suppress distracting information we analyzed the NP
trials. As we did in our previous work (Lupiez and Callejas, 2006, Chapter 1), since
ignored repetition trials were always incongruent and also preceded by an incongruent
trial, to have a suitable condition to compare the results with, only incongruent trials
preceded by incongruent trials were selected for the control condition. A mixed ANOVA
with the factors Group (PSV and Controls) and NP (Control vs. Ignored Repetition) was
carried out on the reaction time data. Means for these conditions can also be found in
Table II and in Graph 7. Again PSV was slower than the control group (F(1, 7)=63.10,
p=.001). No significant NP effect was found (F(1, 7)=1.3, p=.286) and no modulation by
group (F<1).
400
500
600
700
Color Naming Color Naming Phot is m Naming
Cont r ol Gr oup PSV
N
a
m
i
n
g

t
i
m
e

(
m
s
)
Control Ignored Repetition

Graph 7. Experiment 2a and 2b. Negative Priming effect. Reaction times for PSV for the
Color Naming and Photism Naming tasks and Color Naming for her Control.
Reaction time analysis. Experiment 2b: Photismnaming.
Again items were used as random factor. After eliminating the first trial of each
block, a further 2.5% was eliminated due to incorrect triggering of the voice key and
160
1.4% due to errors. A further 1% of trials with extreme values (lower than 200ms and
higher than 1500ms) were also eliminated from the analysis.
Stroop analysis.
First a repeated measures ANOVA for PSV was carried out with N-1
Congruence and N-Congruence as factors. A subsequent analysis comparing PSVs
performance across experiments was carried out. The first analysis showed a
congruency effect of about 40 ms (F
(1, 7)
=15.78, p=.005). This effect was not
significantly modulated by the congruency of the previous trial (F
(1, 7)
=2.30, p=.173),
although there was a trend in the same direction as in the color naming task. Results
are represented in Graph 6 and Table II.
In order to carry out the second analysis a new variable was coded: Experiment
(Color naming vs. Photism naming). We then conducted another repeated measures
ANOVA with the factors Experiment, N-1 Congruence and N-Congruence. A main effect
of Experiment showed that PSV was faster naming the photism than naming the
display color (F
(1, 7)
=20.55, p=.003). The N-Congruence effect was also different across
experiments (F
(1, 7)
=11.51, p=.012). The congruency effect was much smaller for the
photism naming task and this was so because incongruent trials were responded to
faster than in the color naming task (p=.026). Last, N-1 Congruence also modulated N-
Congruence (F
(1, 7)
=12.11, p=.010), and this modulation did not depend on the task
(Experiment). When the previous trial had been incongruent, the current trial showed a
smaller congruency effect due to slower RTs for congruent trials and faster RTs for
incongruent trials.
Negative Priming Analysis
Again we first analyzed PSV for the photism naming task and we then
proceeded to compare her performance on both tasks. The NP ANOVA showed no
effect of NP (F<1). The analysis comparing PSVs NP conditions for both experiments
had two factors: Experiment and NP condition. Again only the main effect of
Experiment was significant (F
(1, 7)
=12.77, p=.009) showing that PSV was faster to name
the photism of a stimulus than the actual color in which it was presented. NP as well as
its modulation by Experiment were not significant (both Fs<1).
161
Accuracy analysis.
Performance was kept very high across Groups and Experiments. Some
conditions showed perfect performance and thus no variance. Therefore, no statistical
analysis was carried out. Percentage correct for each Group and Experimental
condition can be seen in Table II.

We wanted to test whether PSVs synesthesia was not vivid enough as to
produce a congruency effect and whether that was the reason for her not to show any
effect in the emotional task performed in Experiment 1a. Results in the Stroop-NP task
show that PSVs performance does get affected by the match-mismatch between the
presented color and the experienced photism. Even more, as it happens with MA
(Lupiez & Callejas, 2006, Chapter 1), the interference produced by an incongruent
photism, when having to name the ink color, is reliably larger than the interference
produced by an incongruent ink color when having to name the photism. That is, her
patterns of results most resemble that of the projectors shown by Dixon et al. (2004).
It is also interesting to see that PSV is slower than her control group when
naming the ink color. However, if we compare her overall reaction times across
experiments with controls we find that she is not slower than controls when her task is
to name photisms (p=.679). She is as fast as controls to name something that is not
being perceived through her senses but slower than them when having to name the
color that is presented in the screen and visually processed. This also shows that her
experience is quite strong. In fact, she informed several times during the different
testing sessions that she liked the photism naming task more and that she found it
easier to perform than the color naming task. Last, the lack of NP effect shows that
PSV does not succeed in inhibiting the distracting information enough as to make it
less accessible for the next trial.
Results show that PSV does strongly experience photisms and this experience
is sufficiently strong and automatic as to cause an interference with the task to be
performed. Since lack of Stroop interference cannot be put forward to explain her lack
of congruency effect in Experiment 1a we though of a different reason. In the valence
categorization task the dimension of the stimulus she is responding to (emotional
meaning of the words) is orthogonal to the Congruence manipulation (matching
between the presented color and elicited photisms); this could be a factor to explore. In
162
Experiment 1a she did not have to respond with a color name but with a valence
category. Thus she was able to inhibit the irrelevant information and focus on the
required task. In the Stroop task, since the dimension to be ignored shares a response
code with the dimension to be attended to, the interference (i.e. congruency effect)
occurs.
Alternatively, it could be either that since words are more complex stimuli than
letters or numbers, it is easier for PSV to ignore their coloring, or that words elicit
weaker photisms that letters or numbers. Any of these reasons could explain why PSV
does not show a congruency effect in the emotional categorization task with words,
whereas she shows a strong Stroop effect with letters and numbers. Therefore, if color
processing was somehow promoted by the task, we might be able to find some
modulation of color-photism congruency on the emotionality processing of the words or
vice versa. To test this idea we designed a new experiment in which we presented the
same words but now asked her to categorize not the valence of the word, but the
accuracy of their coloring (according to her internal coloring for the words, i.e., her
photisms). The rationale was as follows. Access to word meaning has been shown to
be highly automatic (Mari-Beffa, Fuentes, Catena, Houghton, 2000) and therefore
regardless of whether the response is associated to the meaning or not, it is going to
be processed. However, it might be the case that photisms can be easily ignored when
the task is more demanding and they are not pertinent to it. Now, if this is the case,
when processing of the photism is promoted we would expect it to interact with the
semantic connotation of the word and thus show the typical inversed congruency effect
for negative words. In order to test this hypothesis we conducted Experiment 3.
EXPERIMENT 3. CATEGORIZATION OF COLOR ACCURACY
We designed this experiment to test the hypothesis that photisms elicited by
words (not only those elicited by letters and numbers) do influence PSVs performance
when they are processed, and furthermore, that this influence is modulated by the
valence of the stimulus being processed. In this new task the same set of emotional
words and some additional neutral filler words were presented in both congruent and
incongruent colors (according to the elicited photisms). We asked PSV to discriminate
whether the words were correctly colored or not. Therefore now the color was the
attended dimension of the stimulus and the connotation (the valence of the emotional
words) the ignored one. We also asked MA to run this experiment to compare her
performance with that of PSV. Since MA did show a behavioral effect due to her
163
affective reaction when faced with incongruently colored words being color not a
relevant dimension of the stimulus, we expected her to also show the same
congruency effect here.
METHOD
Participants
PSV and MA run this experiment.
The same set of 26 emotional words used for PSVs Experiment 1 was used
here. An additional set of 13 filler neutral words were also introduced to divert attention
from word valence.
A set of 30 emotional words previously used in MAs experiments were used
here. None of them elicited a black or grey color. An additional set of 15 neutral words
were also introduced to diver attention from word valence.
Since words were chosen so that they did not elicit either black or grey photism,
MAs and PSVs word sets were somewhat different.
Experiment 3 was very similar to PSVs Experiment 1 (this chapter) and MAs
categorization experiments. The variables being manipulated were the Valence
(Positive, Neutral and Negative) of the word and the Congruence between the
presented color and induced photism. Therefore a new level of the Valence variable
was added. Again, congruent colors were those chosen by each synesthete for each
word and incongruent colors were the color-wheel opposites.
PSV run 8 blocks of 78 trials each (39 congruently colored words and 39
incongruently colored words). MA run 4 blocks of 90 trials each (45 congruently colored
ones and 45 incongruently colored ones). She run the experiment twice.
At the beginning of each trial a white fixation point appeared on a grey
background for 1000ms. It was then followed by the target word that was kept on the
screen until subjects response. Synesthetes were to respond by hitting the z or the
164
m key in the keyboard, depending on whether the word was correctly or incorrectly
colored. After response the next trial began.
Auditory feedback was given for wrong responses. The experiment was run in a
dimly lit room.
Reaction time analysis.
Data per Synesthete and experimental condition can be seen in Table III. Trials
were averaged for each item (i.e. word) and items were used as random factor. For
MA, trials were averaged across sessions. Incorrect responses were eliminated from
analysis. These accounted for 1% of PSVs data and 5% of MAs data. A further 2
standard deviation filter was used to eliminate extreme values. A mixed 2 (Synesthete)
x 3 (Valence) x 2 (Congruence) ANOVA was performed. Since words presented to PSV
were not the same as those presented to MA, Synesthete was a between item variable
as well as Valence, whereas congruency was a within items variable.
Table III. Experiment 3. Mean categorization times per synesthete and condition. Error
rates are shown in parenthesis
Valence
Positive Negative Neutral
Subject Congruent Incongruent Congruent Incongruent Congruent Incongruent
PSV
582.0
(1.0%)
614.4
(0.0%)
590.8
(0.0%)
563.0
(0.5%)
582.6
(1.0%)
559.5
(1.9%)
MA
619.6
(5.0%)
649.7
(8.3%)
638.9
(6.8%)
623.0
(0.9%)
640.4
(3.3%)
647.6
(2.5%)

Reaction times did not change as a function of Valence (F<1) or Congruence
(F<1) but they were different for PSV and MA (F
(1, 78)
=44.96, p=.001). MA was slower
than PSV. A modulation of Congruence by Valence was found (F
(2, 78)
=4.40, p=.015). As
shown by planned comparisons (see Graph 8), the color of a positive word was
categorized faster if it was congruent (F
(1, 78)
=5.67, p=.019) and the color of a negative
word was categorized faster if it was incongruent although this was not statistically
reliable (F
(1, 78)
=2.77, p=.100). For neutral words no differences were found between
congruent and incongruent trials (F<1). This pattern was not different for PSV and MA
(F<1).
165
500
550
600
650
700
Negative Neutral Positive Negative Neutral Positive
PSV MA
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T

(
m
s
)

Graph 8. Experiment 3. Reaction times for PSV and MA as a function of word Valence and
Color Congruence.
Accuracy analysis. Experiment 3a
A mixed 2 (Group) x 3 (Valence) x 2 (Congruence) ANOVA was carried out on
the error rates and we found that MA was more error prone than PSV (F
(1, 78)
=10.23,
p=.002). Other than that, no differences were found for Valence, Congruence or the
interaction although the pattern of errors for MA resembled that of the RT analysis.

Experiment 3 shows that PSVs performance does not differ from that of MA
when the processing of the color in which the words are presented is enhanced. In this
case, she does find it easier to categorize positive words when they are colored right
and negative words when they are colored wrong. We therefore interpret PSVs results
as showing that when the affective reaction produced by the perception of a colored
word goes in the same direction as the valence of the word itself, she finds it easier to
respond than in the non-corresponding conditions. Her subjective reports confirming it
is unpleasant to look at words when wrongly colored points to the possibility that the
affective reaction is always present even though it might not always show up
behaviorally.
An alternative explanation, however, to the pattern of results observed for both
PSV and MA could be that they are faster when the response-valence mapping is
consistent (i.e. Yes-Positive or No-Negative) than in those cases when the required
response is yes but the word valence is negative or when the response is no but the
word valence is positive. This effect could be a general one and thus not related to the
166
emotion elicited by the color-photism congruence. In order to test this hypothesis we
conducted the following control experiment. Since non-synesthetes could not perform
this task, as the coloring of the words does not carry additional meaning for them (i.e.
there is no photism to compare with) we designed a modified version in which control
participants also had to discriminate whether the same words were correctly colored,
although based on an arbitrary criterion.
EXPERIMENT 4. CONTROL EXPERIMENT FOR EXPERIMENT 3
Word color for synesthetes is not determined by word connotation but neither is
it by each letters color. It is usually, at least in our synesthetes, the word as a whole or
a competition between some of the letters in the word that are reported to cause the
specific color and not other. Therefore, in order to promote the same type of processing
for controls, we chose a criterion for them to judge when a word was colored right or
wrong that ensured they would access their semantic meaning while not having to pay
attention to the words connotation. Participants were instructed that abstract words
should be colored in cold colors (i.e. blue and purple hues) while concrete words
should be colored in warm colors (i.e. red and orange hues). Their task was to respond
whether the words presented were colored right or wrong according to that criterion.
Therefore, although apart from the concretness dimension these words also had a
connotative meaning, that was irrelevant to the task at hand just as it was in
Experiment 3.
METHOD
Participants
A group of 14 control participants run Experiment 4. Control participants were
Psychology undergraduate students. Their mean age was 21 and all but one were
females.
Stimuli, material, design and procedure
Experiment 4 was as Experiment 3 for PSV except for the colors in which the
words were presented and the instructions given to participants.
Five hues ranging from pure blue to violet color were used as cold colors and
another five hues ranging from pure red to pure yellow were used as warm colors.
167
Colors were assigned to words in a random order and some colors were more common
than others (each color was used from 4 to 11 times in each block) in order to make
sure the variety and frequency of colors was as similar as possible to that of the
synesthetes experiments. Each word maintained its assigned congruent and
incongruent color across blocks. Therefore, in this experiment Congruence referred to
the relationship between the words concreteness and the color in which it was
presented given the coloring criteria put forward. Then a congruent trial would be the
one where participants had to respond yes whereas incongruent trials would be those
were participants had to respond no.
The same analysis was carried out for control participants. Therefore trials were
averaged for each item (i.e. word) and items were used as random factor. Since this
task resulted much more difficult for controls than the task synesthetes were performing
we only analyzed the last 2 blocks to ensure a good amount of practice and therefore
that our results would not be due to the novelty and difficulty of the task. Incorrect
responses (12%) were eliminated from analysis. A further 2 standard deviation filter
was used to eliminate extreme values. A mixed 3 (Valence) x 2 (Congruence) ANOVA
was performed. Mean reaction times can be seen in Table IV and Graph 9.
Table IV. Experiment 4. Mean categorization times per condition. Error rates shown in
parenthesis

Valence
Positive Negative Neutral
Congruent Incongruent Congruent Incongruent Congruent Incongruent
905.3
(7.1%)
996.3
(13.2%)
894.5
(8.2%)
971.9
(7.1%)
880.4
(14.8%)
1000.5
(19.0%)
Words reaction time did not vary as a function of Valence (F<1) but it did as a
function of Congruence (F
(1, 36)
=29.58, p=.001). As shown in Graph 9, congruent words
were responded to faster than incongruent words. That is, saying yes was faster than
saying no, a general finding in most tasks. However this Congruence effect was not
modulated by the Valence of the word (F<1), so saying yes when the word was
positive was not faster than when the word was negative. Similarly, saying no when
the word was negative was not faster than when it was positive.
168
500
600
700
800
900
1000
1100
Negative Neutr al Positive
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T

(
m
s
)

Graph 9. Experiment 4. Reaction times for the Control group as a function of word
Valence and Color Congruence.
Accuracy analysis.
The mixed 3 (Valence) x 2 (Congruence) ANOVA yielded the following results.
Responding to positive and negative words was more accurate than responding to
neutral ones (F
(2, 36)
=11.11, p=.001) and responding to congruent words was marginally
more accurate than to incongruent ones (F
(1, 36)
=4.07, p=.052). However this two
variables did not interact (F
(2, 36)
=2.03, p=.146).
The main aim of this control experiment was to check that the results found in
the previous experiment were not due to factors other than the relationship between
words valence and color-photism relationship. The fact that control participants did not
show an interaction between Valence and Congruence rules out the possibility that the
pattern we found en Experiment 3 was due to responding no being easier if the word
was negative and responding yes easier if the word was positive.
GENERAL DISCUSSION
With this set of experiments we were interested in testing whether the affective
reactions elicited by incongruently colored stimuli found in MA were just an
idiosyncratic feature of her synesthesia or they generalized to other synesthetes. Three
synesthetes were tested for this effect in the first experiment. We presented them with
emotional words both in their right color and in a wrong color, according to their
photisms for each word. By having synesthetes categorize the words according to their
semantic valence we found that, although the three of them reported unpleasantness
associated to the presentation of incorrectly colored words, only two of them replicated
169
the previously found pattern of results. EL showed a modulation of Congruence by
word Valence in RT measures and error rates. PR showed a modulation in RTs but not
in error rates. It is interesting to see that even when the effect is found, it seems to
range from a reduction to a complete reversal of congruency.
The factors underlying this various degrees are still not understood but it is
important to show that the pattern is found. When there is a correspondence between
the connotation of the word and the affective reaction reported by the synesthete (i.e.
positive word that is congruently colored and therefore produces a pleasant sensation
or alternatively, a negative word incongruently colored and therefore producing an
unpleasant sensation) their performance is relatively better than in conditions where
the result of the connotation assessment goes in the opposite direction as the result of
the synesthetic assessment (i.e. for a positive word incongruently colored the
connotation assessment signals positive and the synesthetic assessment signals
negative. For a negative word congruently colored the connotation assessment
signals negative while the synesthetic assessment signals positive).
Since PSV did not even show a difference between congruently and
incongruently colored words we carried out a Stroop experiment to test whether her
synesthetic experiences were automatic enough as to influence her performance. A
standard synesthetic Stroop showed that she did find it more difficult to name the ink in
which a letter or number was presented when the elicited color did not match the ink
color. Even more, in a reversed Stroop task in which she had to ignore the ink color of
the stimulus to name the in her-minds-eye experienced photism the interference
produced by the color presented in the screen was much smaller. It seems then that
she did perform in a similar manner as other synesthetes like MA (Lupiez & Callejas,
2006, Chapter1). Even more, based on her pattern of Stroop results it could be
concluded that her synesthetic experiences are of a relative strong degree (strong
enough to influence visually perceived colors and strong enough to be relatively
uninfluenced by them). Adding to this, the absence of a Negative Priming effect on
neither of the two tasks suggests that she was not successful in the suppression of the
interfering information.
Having found a strong pattern of Stroop interference we went back to the
affective side effect of synesthesia and tested whether the absence of a congruency
effect in this realm was due to the fact that color was an irrelevant dimension of the
stimulus and since the task was more demanding, color processing was taken apart as
to not influence PSVs performance in the valence categorization task. By making her
170
categorize whether the color matched or mismatched the words elicited photism we
ensured that she carried out enough processing on the color as to experience the
reported unpleasantness associated with wrongly colored words. Results showed that
when color processing was promoted, correspondence between the affective reaction
induced by the color-photism relation and the connotation of the word influenced
performance and made her respond faster and more accurate to corresponding
conditions than to non-corresponding conditions. The fact that MA showed the same
pattern of results backs up the deduction that the affective reactions underlying
performance in Experiment 1a (where MA also showed an effect, see Callejas et al.,
submitted, Chapter 3) are the same ones playing a role here. Even more, the fact that
control participants did not show an effect in the same direction supports the idea that
the modulation found is not due to the fact that responding no to negative words and
yes to positive words is easier than the other conditions.
The question to be answered now is why, given PSVs subjective reports and
her clear patterns of interference found in the Stroop tasks, she did not show an effect
in Experiment 1a. Even though color was completely irrelevant in this experiment, the
task was not as resource demanding as to prevent automatic processing of the color. In
fact, she did report realizing some words were not colored right. Therefore she was
conscious of the mismatch between color and photism. It could be the case that her
synesthesia for words is of a more higher level (Ramachandran & Hubbard, 2001b)
and connotation assessment was already carried out by the time the photism was
experienced and the unpleasantness reaction was felt.
Alternatively other personality factors could be playing a role in PSVs
performance. Some factors such as anxious traits could be enhancing the affective
reaction so much as to make it influence performance (Acosta, Lupiez, Fresse &
Megas, under revision). When testing PSV and MA with the Spanish normalized
version of the STAI test to measure trait-anxiety, we found that PSVs scores were in
the lowest percentiles (score:11, percentile: 11) while MAs were in the highest ranges
(score:34, percentile: 80). Although no cause effect can be drawn from this data, it
informs of individual differences other than the synesthesia experience per se that
could be influencing, not only these affective aspects of synesthesia but many other
aspects of their perception and experiences. It could be that having a low anxiety trait
helps in coping with the anxiety state elicited by the incongruence situation so that it
does not influence performance. This would not be so when having a high anxiety trait.
However, the incongruence situation presented in Experiment 3 is much more subtle
171
and it may not be actively dealt with resulting in it being observable in MA as well as in
PSV. In fact, both synesthetes reported liking this experiment much more than the
previous ones. The fact that their task was to judge whether something was colored
right or wrong placed them in the position to have control over the situation and not
have to disregard their experiences in order to perform an unrelated task (as was the
case in Experiment 1). These observations encourage researchers to broaden the
scope of the circumstances taken into account when studying synesthesia and
specially when trying to discern the underlying factors that cause the differences
between synesthetes in their performance on behavioral tasks.

CHAPTER 5
CONTEXT EFFECTS ON SYNESTHETICALLY
INDUCED AFFECTIVE REACTIONS

Chapter 5. Context Effects in Synesthetically Induced Affective Reactions _____________
175
ABSTRACT
Informal reports from synesthetes suggest that the affective reaction
experienced when perceiving an incongruently colored stimulus might be permeable to
contextual influences. In order to test these reports we carried out a set of four
experiments where we manipulated the color in which a set of words were presented
(i.e. black or grey) as well as the context in which they were shown (i.e. achromatic vs.
chromatic context) and tested how these factors influenced our synesthetes ratings for
the valence of those words. Also we tested how the same shade of grey would affect
their performance when it was perceived as white. We found that black was clearly
susceptible to context effects whereas perception of grey tended to be unpleasant
under chromatic and achromatic conditions. Last, grey colored-white perceived words
seemed to be less incongruent in color than the very same stimuli when subjectively
perceived as grey. Results show that the affective reaction elicited by the incongruence
between the presented color and the elicited photism is not driven exclusively by the
stimulus but also influenced by other factors.
INTRODUCTION
Synesthesia is a rare phenomenon whereby the conscious perception of a
stimulus is accompanied by the subjective experience of another sensation not directly
linked to it (Grossenbacher, 2001). Being grapheme-color synesthesia the most
common case (Day, 2005), it is also the one most deeply studied to the moment
(Grossenbacher, 2001; Ramachandran & Hubbard, 2001b; Rich & Mattingley, 2002 for
reviews).
Previous informal reports have pointed to the emotional nature of synesthesia
(Cytowic, 1993; Ramachandran & Hubbard, 2001b). Synesthetes experience their
photisms as a vivid phenomenon, which is accompanied by a sense of certitude that
what they see is real. Synesthetes also report experiencing a feeling of disconfort when
a stimulus is presented in a color that does not match the internally experienced
photism. Recently, Callejas, Acosta and Lupiez (submitted, Chapter 3) showed for
the first time that this subjective feeling of disconfort can be objectively measured and
reproduced in a laboratory setting. Even more, it has been shown to be automatic and
difficult to ignore when it interferes with the task being performed.
___________________________________SECTION 2. SYNESTHESIA AND EMOTIONS
176
Informal reports from synesthetes lead us to think that although automatic in
nature, these affective reactions might be permeable to contextual influences. Several
synesthetes have reported that although acknowledging that something is wrongly
colored, the incongruence can be ignored if necessary. Particularly, when reading
books or handwriten notes, they do not seem to be so affected by the ink color in which
the text is writen. Most synesthetes inform that black colored items are not as
disrupting to see as any other color.
Previous reports have informed that the experience of synesthesia itself can be
modulated by the context in which the elicitor is placed (Dixon et al., 2006; Myles,
Dixon, Smilek & Merikle, 2003). These researchers showed that the very same
perceptual shape could be experienced as a letter S or a number 5 depending on
the context in which it was presented (i.e. HPSOAor 37564). By combining the color
elicited by each grapheme (i.e. by the letter or by the number) with the context in which
it was presented (i.e. letter context or number context) they showed that the very same
perceptual shape with the very same color could give rise to color naming times very
different for congruent than incongruent conditions (up to 300ms congruency effect).
Given this findings whereby the strong vivid experience of a color can be
manipulated by the contextual cues that point to the identity of ambiguous grapheme
shapes, we wanted to test whether the affective reaction resulting from the perception
of an incongruently colored item could be also manipulated by the context in which the
eliciting stimulus was presented. If a word is presented in a context in which all words
are uniformly colored, even though not in their right color, it is possible that synesthetes
would interpret that as a color-independent context and therefore try to avoid or control
the affective reactions elicited by the incongruence in order to perform better (i.e. easily
read a posted letter written by a friend with a blue ink ball pen).
In order to do so we systematically manipulated the color, and context in which
the color was presented, of a set of neutral and emotional words.
Previous experiments were we had asked our synesthete MA to make a
valence judgment of a set of words, showed that when presented with colored words
her ratings were influenced by the congruency between the words color and the
photism induced by the word (Callejas et al., submitted, Chapter 3). As can be seen in
Graph 1, her ratings were overall more positive when the word was presented in the
congruent color and more negative when the word was presented in an incongruent
color as compared to a control group.
177
-3
-2
-1
0
1
2
3
N
e
u
t
r
a
l
P
o
s
i
t
i
v
e
N
e
g
a
t
i
v
e
A
n
x
i
e
t
y
N
e
g
a
t
i
v
e
A
n
g
e
r
N
e
u
t
r
a
l
P
o
s
i
t
i
v
e
N
e
g
a
t
i
v
e
A
n
x
i
e
t
y
N
e
g
a
t
i
v
e
A
n
g
e
r
Controls M.A.
V
a
l
e
n
c
e

J
u
d
g
e
m
e
n
t
s

Graph 1. Results from Experiment 1a and 1b of Callejas et al. (submitted, Chapter 3).
Error bars denote standard error of the mean.
We also found, as shown in Graph 1, that even though black was not the
photism of any of the presented words, and therefore it is an incongruent color, when
shown the very same words but now all colored in black, MA presented the same
pattern of results as control participants did (compare black bars in the right side of
Graph 1 with grey and white bars in the left side). That is, she correctly evaluated the
words with no influence from the color congruency when they were consistently colored
in black, ad it is usual in normal reading texts.
EXPERIMENT 1. EFFECT OF COLORING WORDS IN BLACK IN A COLOR-
CONTEXT TASK
To test whether black is in fact considered a neutral color or the results we had
found were due to the fact that it was presented in an achromatic context (grey
background and all items presented in black) we carried out our first experiment where
we presented the same words to MA and a group of control participants, but now in one
of three different colors, randomly selected: congruent, incongruent and black. The
rationale was the following: if absence of hue is the reason for black to be considered
neutral, then we should still find an accurate assessment of black colored words when
presented in a color context (i.e. other colors are intermixed with black). However, if the
context is the determinant for a black word to be perceived as neutrally colored then
we would expect to find the same pattern of results for black than for the incongruent
color condition. After all, a black colored word is an incongruently colored word if black
is not the photism elicited by such word.
178
METHOD
Participants
MA and a group of 10 undergraduate psychology students participated in this
experiment. They all had normal or corrected to normal vision and the mean age of the
group was 20. All of them were women (including MA).
Out of the 72 words used in the Callejas et al. study, a set of 58 were selected
so that neither of them elicited a black or a grey photism. A total of 15 positive words,
15 neutral words and 28 negative words (14 anger related and 14 anxiety related) were
selected.
Words were presented on a grey background screen where a white
reproduction of the response scale was present through out the whole experiment.
The 58 words were presented in one of three different color categories
depending on the condition. Congruent colors were those chosen by MA for each
particular word. Incongruent colors were the color-wheel opposites of the congruent
ones. Last, words were also presented in black color. To ensure good viewing
conditions, a dimly lit room was used to carry out the experiment.
Participants task was to evaluate the valence of each word in a 7-point Likert
scale being -3 very negative, +3 very positive and 0 neutral. A speeded response was
not requested although participants were encouraged not to engage in long decision
processes. Thus, we manipulated Congruency (congruent, incongruent and black) and
Valence (positive, neutral, anxiety related and anger related negative words).
On each trial, a white fixation point was presented for 1000ms prior to the
appearance of the target word, right above the reproduction of the rating scale (-3 -2 -1
-0 +1 +2 +3). After subjects response the target word disappeared and the next trial
begun. The response scale was present at all times.
Four blocks of trials were run and a total of 174 trials were included in each
block (58 words x 3 different coloring conditions). Conditions were randomly
179
intermingled in each block. Prior to the experimental blocks, a set of practice trials was
run. Participants were encouraged to rest between blocks.
For each group data was averaged for each condition and item. Items and not
subjects were used as random factor. Therefore, data for the control group was
averaged across participants for each item. Group and Congruency were within item
factors and Valence was a between item factor. Mean rating for each group and
condition can be seen in Table I.
Table I. Experiment 1. Mean valence judgment per group and condition.
Valence
Group Congruence Positive Negative-Anxiety Negative-Anger Neutral
Congruent 2.32 1.54 2.09 1.93
MA Incongruent -1.50 -2.11 -1.91 -1.72
Black -0.77 -0.91 -1.07 -0.35
Congruent 2.21 - 1.95 - 2.19 0.18
Incongruent 2.25 - 1.98 - 2.13 0.13 Controls
Black

2.23 - 1.94 - 2.22 0.18

The mixed 2 (Group) x 3 (Congruence) x 2 (Valence) ANOVA yielded the
following results. Words were correctly rated according to their meaning (F
(3, 54)
=186.8,
p=.001) although there was also a significant interaction by group (F
(3, 54)
=72.8,
p=.001). MAs ratings were more similar for all word types. However, planned
comparisons revealed that she did rate positive words more positive than negative
words (F
(1, 54)
=46.52, p=.001 and F
(1, 54)
=6.5, p=.013 for anxiety related and anger
related words respectively) although neutral words were not different from positive
words (F<1). Groups were not different (F
(1, 54)
=2.20, p=.140) although the Congruence
effect found (F
(2, 108)
=227.61, p=.001) was modulated by Group and Valence (significant
second order interaction: F
(6,108)
=5.91, p=.001). As can be seen in Graph 2 control
participants did not differ in their ratings when words were colored or presented in
black. MA showed a very intriguing evaluation pattern. A separate analysis of her data
showed that she rated black words significantly higher than incongruently colored
words (F
(1, 54)
=42.26, p=.001) but much lower than positive words (F
(1, 54)
=181.18,
p=.001).
180
-3
-2
-1
0
1
2
3
Anxiet y
Negat ive
Anger
Anxiet y
Negat ive
Anger
Cont r ols M.A.

Graph 2. Experiment 1. Valence rating for congruent, incongruent and black colored
words as a function of group and valence. Error bars denote standard error of the mean.
An interaction by Valence was found in MA (F
(6, 108)
=5.69, p=.001) and was due
to the fact that valence judgments were more extreme when words were colored
congruently and specially due to negative words related to anger being rated positive
when congruently colored. Although puzzling and unexpected, this could be accounted
for if we take into account MAs high anxiety-trait (see Chapter 4) that would enhance
the processing of anxiety related negative words and her positive mood at time of
testing that might have reduced the relative connotation of anger related words in the
context of congruently colored words. More testing is needed however to decipher
these results.
In is interesting to see that, although overall MA is following instructions to rate
words according to their semantic valence, she is still influenced by the congruency
between the words connotation and the photism-color match. Overall, results from
congruently and incongruently colored words resemble those found in previous
experiments (Callejas et al., submitted, Chapter 3). Black colored words on the
contrary, were rated much more negatively than congruently colored words. In fact,
ratings of black colored words were close to those of incongruently colored words, and
therefore quite different of control group ratings. Whereas MA rated black colored
words exactly as control participants did when black was constant throughout the
experiment (Callejas et al., submitted, Chapter 3). It therefore seems that what is seen
as a neutral word under no coloring conditions (i.e. black words in an achromatic
context) becomes an incongruent color under chromatic conditions.
181
We then compared the results of MAs original black experiment (Experiment 1b
in Callejas, et al. submitted, Chapter 3) with those found here for the very same words
in the very same color. Since this new experiment had eliminated some words that
elicited a grey photism, those items were also eliminated from the original experiment
prior to comparison. A new variable, called Context (achromatic vs. chromatic context),
was coded. Graph 3 shows the results of this analysis. Valence rating was influenced
by Context (F
(3, 54)
=90.86, p=.001). Post-hoc tests showed that while positive words
were rated more positive in an achromatic context (p=.001), negative words were rated
more negative (p=.006 for anxiety related words and p=.001 for anger words). No
differences were found for neutral words (p=.292). Therefore, words rating was more
accurate in the achromatic condition. This was also shown in the original paper
(Chapter 3) when black was shown to be different than MAs performance under color
conditions but not different from Controls performance under color conditions.
-3
-2
-1
0
1
2
3
Neut r al Pos it ive Negat ive Anxiet y Negat ive Anger
Achromatic Chromatic

Graph 3. Experiment 1 compared to Experiment 1b from Callejas et al. (submitted,
Chapter 3). Ratings for black colored words as a function of color Context and Valence.
It seems to be clear then that context plays an important role and the very same
word presented in the very same color can influence behavior in one direction or
another depending on the context in which it is found. Whether context can also
influence the affective reaction itself (its presence or strength) will be discussed in a
later section.
In this experiment we wanted to test whether black is a universal neutral color
or it is only perceived as so when it is presented in an achromatic context. Having
found that black behaves differently depending on the context, our next interest was to
discern what feature of a black word presented in an achromatic context makes it turn it
into a neutral colored word. Considering black is absence of any hue and grey is just a
182
brighter version of black, we wondered whether absence of hue in an achromatic
context would be the determinant of neutrality irrespective of brightness. In order to test
this idea we conducted experiments 2a and 2b.
EXPERIMENT 2a AND 2b. EFFECT OF COLORING WORDS IN GREY IN A
COLOR-CONTEXT TASK AND AN ACHROMATIC-CONTEXT TASK
We next wanted to test whether grey would behave just as black since it is a
brighter version of black. With this new manipulation we wanted to further test whether
black was considered a neutral color because it lacked a hue or just because it is the
most used color when reading printed documents and therefore, the incongruent color
that synesthetes are most used to deal with. If the first reason was correct, then grey or
white would behave in a similar manner. However, if the reason is extensive practice
reading black print, then we would expect to find that grey would be considered an
incongruent color irrespective of its context. Again we run the same experiment but
now words were presented either in a congruent, incongruent or grey color (Experiment
2a) or just in grey color (Experiment 2b).
METHOD
Participants
Experiment 2a. MA and a different group of 10 undergraduate psychology
students participated in this experiment. They all had normal or corrected to normal
vision and the mean age of the group was 22. Two of them were men.
Experiment 2b. MA and the same group of participants that performed
Experiment 1 participated in this experiment.
Experiment 2a. Since the background was dark grey, we chose a lighter shade
of grey for the words (RGB: 180,180,180). Congruent and incongruent words had the
same color as in Experiment 1 and we replaced black words with the grey words.
Everything else was the same as in Experiment 1.
Experiment 2b. Words were presented only in grey color and the shade of grey
used was the exact same one used in Experiment 2a. The same number of trials was
183
presented but since now only one color was used, three times as many trials as in
experiment 2a were collected for each condition. Everything else was the same as in
Experiment 1 and 2a.
The grey shade used for the background and the grey colored words, as well as
the procedure, is shown in Figure 1.

+
-3 -2 -1 0 +1 +2 +3

Muerte
-3 -2 -1 0 +1 +2 +3
Figure 1. Experiment 2a and 2b. Stimuli and procedure used.
MA run Experiment 2a twice and Experiment 2b only once. Control participants
run the experiments only once.
Experiment 2a.
Again, for each group, data was averaged across participants (across sessions
for MA) for each one of the 58 items. Mean valence judgments for each Group and
Condition can be seen in Table II. The mixed 2 (Group) x 3 (Congruence) x 2 (Valence)
ANOVA yielded the following results.
184
Table II. Experiment 2a. Mean valence judgment per group and condition.
Valence
Congruent 1.71 0.47 0.86 1-32
Incongruent -0.43 -1.05 -1.13 -0.75 MA
Grey -1.67 -1.67 -1.7 -1.58
Congruent 1.81 - 1.80 - 1.95 0.29
Incongruent 2.00 - 1.74 - 2.07 0.28 Controls
Grey 1.82 - 1.67 - 1.91 0.28

Participants followed instructions and negative words were rated more negative
than positive words being neutral words in between (F
(3, 54)
=117.39, p=.001). Groups
were not different (F
(1, 54)
=2.36, p=.130) although they did differ in their Valence
judgments (F
(3, 54)
=65.77, p=.001). Again MAs ratings were more uniform than those of
the control group (see Graph 4). However, planned comparisons showed that she rated
positive words more positive than negative ones (F
(1, 54)
=14.19, p=.001 and F
(1,
54)
=10.21, p=.002 for anxiety related and anger related words) and those more negative
than neutral ones (F
(1, 54)
=6.28, p=.015 and F
(1, 54)
-3
-2
-1
0
1
2
3
Anxiet y
Negat ive
Anger
Anxiet y
Negat ive
Anger
Cont r ols M.A.
Congruent Incongruent Grey

Graph 4. Experiment 2a. Valence rating for congruent, incongruent and grey colored
words as a function of group and valence. Error bars denote standard error of the mean.
The effect of Congruence (F
(2, 108)
=163.55, p=.001) was modulated by Group
(F
(2, 108)
=188.10, p=.001) and also by Valence (F
(6, 108)
=2.37, p=.035). While control
participants did not show any difference as a function of the color in which the words
were presented (all ps>.9), MA showed that congruently colored words were the most
positively rated ones, incongruent words were more negative and grey words were the
most negative ones (all ps<.001). Inspection of the modulation of Congruence by
185
Valence showed that the interaction could be due to the congruency effect (i.e.
difference between congruent and incongruent trials) being larger for positive, neutral
and anger related negative words than the anxiety related negative words. Obviously,
although the second order interaction was not significant, since control participants had
not shown a congruency effect in the first place, it was clear that this interaction
between Congruence and Valence was due to MAs ratings. Since we predicted MA to
perform differently at least between congruently and incongruently colored words we
carried out a separate ANOVA for her.
The 4 (Valence) x 3 (Congruence) ANOVA showed a main effect of Valence (F
(3,
54)
=6.07, p=.001) with the expected pattern and a main effect of Congruence (F
(2,
108)
=191.70, p=.001). Not only congruent words were rated more positive than
incongruent ones but also grey words were rated more negatively than incongruent
ones (all ps<.001).
Experiment 2b
Averaged data per group can be seen in Table III. A mixed 2 (Group) x 2
(Valence) ANOVA was performed on the data.
Table III. Experiment 2b. Mean valence judgment per group and condition.
Valence
MA Grey - 0.71 - 2.89 - 2.84 - 0.90
Controls Grey 2.07 - 1.96 - 2.18 0.13

Results showed that words were assessed accurately as positive words were
rated higher than neutral and both higher than negative ones (F
(3, 54)
=348.26, p=.001).
Groups were different so that MA tended to rate all words more negative than the
control participants did (F
(1, 54)
=235.72, p=.001). An interaction was found between
Valence and Group (F
(3, 54)
=32.19, p=.001) in the following direction. As can be seen in
Graph 5 although MA was in general more negative in her ratings than the control
group, this difference was maximized for the positive words. Post-hoc analysis showed
that she did not rate positive words statistically different than neutral words (p=.9). All
other comparisons were significant (all ps<.01).
186
-4
-3
-2
-1
0
1
2
3
Neutral Positive Negative
Anxiety
Negative
Anger
V
a
l
e
n
c
e

J
u
d
g
e
m
e
n
t
s
Controls MA

Graph 5. Experiment 2b. Mean valence judgments for MA and the Control group when
words were all presented in grey color. Error bars denote standard error of the mean.
In this experiment we also studied the valence of a control group with grey color
words because the original studies used to estimate mean valence of each word
(unpublished data) had been carried out with black presentation and we wanted to
make sure any potential effect found in MA was not due to the color itself but to its
interaction with the experienced photism. Having found the already mentioned results,
we can conclude that grey color on its own was not the reason for MAs pattern since
control participants rated words just as they did in the original study and in the previous
studies where other colors were used (Experiment 1 and Experiment 2a from this
chapter and Experiment 1a from Callejas et al., submitted, Chapter 3).
In order to study the relative effect of context in MAs responses to grey colored
words we carried out another analysis comparing her performance for grey words
under an achromatic and a chromatic context. Therefore we had a new variable
Context with 2 levels (achromatic vs. chromatic). Results are shown in Graph 6. Apart
from the main effect of Valence (F
(3, 54)
=63.12, p=.001) we just found a trend for a main
effect of Context with chromatic context tending to produce more positive ratings than
achromatic context (F
(1, 54)
=3.42, p=.070). This effect was modulated by Valence (F
(3,
54)
=40.17, p=.001). When grey was presented in an achromatic context judgments were
more accurate. Negative words were rated as more negative than neutral ones (both
ps<.001) and more negative than positive ones (both ps<.001) although there was no
difference between neutral and positive words (p=.858). However, in the chromatic
context all words were rated as negative irrespective of the semantics of the word
(mean -1.66).
187
-3
-2
-1
0
1
2
3
Neut r al Pos it ive Negat ive Anxiet y Negat ive Anger
Achromatic Chromatic

Graph 6. Experiment 2a and Experiment 2b for MA. Ratings for grey colored words as a
function of color Context and Valence.
To further study the effect of color and context we carried out another ANOVA
for MA in which we included all the achromatic conditions of the 4 experiments. We
then had three variables: Color (black vs. grey), Context (achromatic vs. chromatic)
and Valence (positive, neutral, anxiety and anger related negative). As shown in Graph
7, we found that grey words were rated as more negative than black words (F
(1,
54)
=253.17, p=.001). Although the effect of Context was not significant (F
(1, 54)
=2.32,
p=.134), it did modulate Color (F
(1, 54)
=20.53, p=.001). Again planned comparisons
showed that black was highly influenced by context (F
(1, 54)
=18.05, p=.001) whereas
grey only tended to be influenced by context (F
(1, 54)
=3.42, p=.070).
-3
-2
-1
0
1
2
3
Anxiet y
Negat ive
Anger
Anxiet y
Negat ive
Anger
Achr omat ic Chr omat ic
Black Grey

Graph 7. Comparison of results from Experiment 1b (Callejas et al., submitted, Chapter
3), Experiment 1, 2a and 2b. Ratings for Black and Grey words as a function of Valence
and color Context.
The unexpected pattern of ratings found for grey when shown in a color context
(i.e. all words were rated as negative irrespective of their valence) caused some other
188
interactions to be significant. Valence was influenced by Context (F
(3, 54)
=105.65,
p=.001) in the sense that ratings were more accurate under achromatic context
conditions. In the chromatic context, ratings were more homogeneous for all valence
types mainly due to the effect of grey. Valence was also influenced by Color (F
(3,
54)
=18.25, p=.001). Although grey words were consistently rated as more negative than
black words, this difference was maximized for positive words. Graph 7 clearly points to
the source of this interaction. Black positive words under an achromatic context were
the only positively rated words.
The second order interaction between Valence, Color and Context was reliable
(F
(3, 54)
=13.22, p=.001). Ratings for words presented in an achromatic context followed
the expected valence pattern. However, when shown in a chromatic context, ratings
were much more undifferentiated across valence conditions and this was especially
true in the case of grey.
These results suggest that context affects differently to grey and black words.
While black words act as incongruent colors when in a chromatic context and as a
neutral color when in isolation, grey is only marginally influenced by context and shows
up as an incongruent color independent of context. Results from Callejas et al.,
(submitted, Chapter 3) showed that black in achromatic context did not differ from a
non-synesthetes rating. Here we showed that although grey resembled the expected
pattern of results under achromatic context in the sense of the relative difference
between positive and negative words, it still differed from the control group because all
the ratings were shifted towards the negative values. Grey does have a different status
than black. It is worth mentioning that, when grey was presented in an achromatic
context the actual rating values were quite similar to those originally given by MA to
incongruently colored words (see Graph 1). Therefore grey seems to be as unpleasant
when showed alone as any other incongruent color. A question rises then concerning
the different patterns found for MA when rating congruently and incongruently colored
words. Why would her ratings change over time? We will come back to this issue in a
later section.
Last, while designing these experiments we realized that we could take
advantage of the fact that the very same color could be subjectively perceived as grey
or white as a function of the other colors presented in the screen to also test how
subjective perception and not wave length affect MAs color preferences. In order to
test this we carried out Experiment 3.
189
EXPERIMENT 3. SUBJECTIVE EXPERIENCE OF COLOR VS. COLOR WAVE
LENGHT
When designing Experiment 3 we noticed that the grey shade chosen to color
the words looked almost white when shown in a context where all other colors were
darker (i.e. darker grey and black). However, the same shade was subjectively
perceived as grey when lighter colors were in scene (i.e. darker grey and white). We
took advantage of this contrast effect to test to a greater extent how color incongruence
can be modulated by subjective perception.
METHOD
Participants
MA and a different group of 10 undergraduate psychology students participated
in this experiment. Their mean age was 21 and all but one were females. MA run this
experiment before she run Experiment 2b
For this experiment, only anxiety related negative words were used. We
therefore had 15 positive words, 15 neutral words and 15 negative words (one anger-
related negative word was used to make groups even in number and subjective
familiarity). The fixation point as well as response scale presented on the screen were
now colored in black ink (instead of white as in previous experiments) promoting the
subjective experience of words being colored white. Everything else was same as in
Experiment 2b.
190
The procedure was the same as in previous experiments. Figure 2 shows how
the subjective perception of the word changes to whitish when set in a black and dark
grey context.

+
-3 -2 -1 0 +1 +2 +3

Muerte
-3 -2 -1 0 +1 +2 +3
Figure 2. Experiment 3. Stimuli and procedure used.
Averaged data per group can be seen in Table IV. A mixed 2 (Group) x 3
(Valence) ANOVA was performed. Again, participants correctly followed instructions
and evaluated positive words more positive than neutral words and negative words
more negative than neutral words (F
(2, 42)
=310.57, p=.001).
Table IV. Experiment 3. Mean valence judgment per group and condition.
Valence
Group Congruence Positive Negative-Anxiety Neutral
MA Grey - 0.07 - 2.22 - 0.18
Controls Grey 2.18 - 2.01 0.01

As in the previous experiment, MA rated words more negative in general than
the control group (F
(1, 42)
=117.64, p=.001). However, this effect depended on the
valence of the word as shown by the significant interaction (F
(2, 42)
=70.26, p=.001).
Again, (see Graph 8) positive words were rated more negative by MA than the control
group (p=.001). It is interesting though to note that this was the only statistical and
nominal difference between her and the control participants. For neutral words as well
191
as for negative words there were no differences between both groups in their ratings
(all ps>.6).
-3
-2
-1
0
1
2
3
Neutral Positive Negative Anxiety
V
a
l
e
n
c
e

J
u
d
g
e
m
e
n
t
s
Controls MA

Graph 8. Experiment 3. Mean valence judgments for MA and the Control group when
words were all presented in subjectively perceived white color. Error bars denote
standard error of the mean.
-3
-2
-1
0
1
2
3
Anxiety
Anxiety
Controls MA
V
a
l
e
n
c
e

J
u
d
g
e
m
e
n
t
s
Grey White

Graph 9. Experiment 2b and 3. Mean valence judgments for MA and the Control group as
a function of Valence and subjectively perceived color. Error bars denote standard error
of the mean.
Last, we carried out another ANOVA comparing Experiment 2b and Experiment
3 in order to study the relative influence of perceived color on MAs ratings. Since
Experiment 3 had been carried out only with anxiety related negative words, we only
took those same words from Experiment 2b. We had a mixed 2 (Group) x 2 (Perceived
Color) x 3 (Valence) design. We found that there was a difference between perceiving
words as grey and perceiving them as white (F(1, 42)=58.88, p=.001). As shown in
192
Graph 9, grey perceived words were rated more negative than white perceived words
but, as shown by the significant interaction (F(1, 24)=86.52, p=.001), this was so only
for MA. Control participants, as expected, did not show any effect of color.
Experiment 3 showed how the very same color was assessed as more or less
pleasant (and therefore influenced valence ratings differently) based on the subjective
color experience. When the very same color is categorized as white, valence ratings
are just rated about 0.7 units less negative than when the very same color is
subjectively categorized as grey (given a brighter context). This is consistent with MAs
reports after performing both experiments. While after having finished Experiment 2
she informed that most of the colors of the words were wrong, after finishing
Experiment 3 she commented that it was not so bad since white is not such a bad
color. This seems to be an interesting demonstration that a) the evaluation of the color-
photism fit is not a set process but one that is affected by different factors and b) not all
wrong colors are equally wrong. Again, the question comes back to which are the
determinants of a color being categorized as wrong and therefore causing an
unpleasant feeling. We now turn to that question.
GENERAL DISCUSSION
The general aim of these sets of experiments was to study the factors involved
in the subjective perception of a color as incongruent by a particular synesthete. In
previous work (Callejas et al., submitted, Chapter 3) presented words had to be rated
according to their semantic meaning and when all words were presented in black our
synesthete did not behave differently than controls. Therefore we found that an isolated
presentation of black words did not produce the usual pattern found for incongruently
colored stimuli.
In Experiment 1 we wanted to test whether black was considered neutral
because it was presented in isolation or because it has some intrinsic features that
make it special in that sense. We found that under color context conditions (i.e., the
words could be displayed either in black or in several colors) black behaved as an
incongruent color although it did not seem as unpleasant as the incongruent colors we
used (the color wheel opposites to the congruent ones).
In order to further test whether black had a special status when presented in
isolation (maybe due to extensive practice with printed books) or the results we found
193
were caused by its absence of hue, we used a brighter version of black (i.e. grey) and
performed the very same experiments. Under color context conditions we found that
grey was assessed as even more negative than the incongruent colors. Also, if grey
was presented in isolation it did not induce the expected neutral or normal rating of
words but one very similar to that found when incongruent colors were used.
Last, in Experiment 3 we compared the very same color shown in isolation with
contextual cues that biased subjects to perceive it as grey or as white. Again we found
that a white perceived color did not behave as a neutral shade but as an incongruent
one although less interfering that grey.
It seems clear that MA was performing the task as instructed and not
responding to the color while ignoring the meaning of the word. Two facts back this up.
On the one hand, although fluctuating, her relative pattern of ratings for positive and
negative words is stable across experiments. Positive words are always rated higher
than negative ones and neutral words fall in between but closer to the rating of positive
ones. This result would be difficult to accomplish on a volitional basis since her
response pattern is a dynamic one that only has the commonality across experiments
of keeping the relative difference between positive and negative valence. On the other
hand, if MA was answering solely based on the color in which the words are presented,
the pattern would look quite different. Since congruent words have the color she chose
for them and incongruent ones the color wheel opposite, there is no room for mistake
and she would consistently rate all congruent words with a 3 and all incongruent words
with a -3. In fact, a similar experiment was carried out, in which we asked MA to
respond whether the words were correctly or incorrectly colored, and she performed
the task perfectly (see Chapter 4).
Thus, once again, in spite of MA responding to the word valence (i.e., according
to instructions) she was influenced by the correspondence between the coloring of the
words and the photisms elicited by them. The results found in these experiments when
presenting the words in an achromatic shade suggest that black is special. Even more,
black is only special under certain circumstances. As soon as a cue leads to the
thought that the context is a chromatic one, black becomes just as incongruent as any
other color. Also, achromatic colors seem to have different status. While black can be
taken as neutral when in an achromatic context, grey is still considered as very
incongruent. A white perceived color on the other hand, seems to be somewhat less
interfering than grey but still more incongruent than black.
194
Now, if synesthetes like MA claim such a strong aversive reaction to wrongly
colored words the question comes to how can they cope with black printed stimuli? Is it
that they loose their synesthesia when the stimuli are presented in black? Is it that they
loose the emotional reaction associated to the synesthesia experience? Experiment 1
showed that the synesthesia experience is not lost, at least when black is in a color
context and neither the affective reaction as measured by the influence of congruency
on the valence ratings. If the photism is still experienced in a color context, it is
presumably also experienced in an achromatic context.
The fact that different samples of achromatic shades elicit different responses
points to the assessment process categorizing stimuli as congruent or incongruent with
the internally experienced phenomena as being flexible. And it seems that one of the
factors that can influence this process is the amount of practice with such stimuli.
Therefore, the fact that black is the most used color for printed books, newspapers and
the like and the fact that usually printed books do not show black print mixed up with
any other colors could explain the results found for black. Given this premises it is
possible that after long practice with the very same words presented in the very same
congruent and incongruent colors, MA has started to habituate to those incongruent
colors and her ratings are not as extreme as would be initially (i.e. in Experiment 1a
from Callejas et al., submitted, Chapter 3). That would explain why grey was rated as
more negative than incongruent colors in Experiment 2b. While this was the first time
MA was dealing with grey words, she had extensive practice with the incongruently
colored words presented in the study. However, this does not explain why white (the
same shade perceived as white) was rated less negative, although she had no practice
rating words in this perceived color.
Extensive questioning to MA suggests that the synesthesia experience is still
there and the affective reaction might be there too but synesthetes undergo an
adaptation process to fit into the environment and prevent their synesthesia from
becoming a challenge which might make them habituate to certain inconsistencies to
better cope with them. This process seems similar to that found in certain phobia
therapies where the behavioral response (i.e. escape) is controlled whereas the
physiological or the cognitive one is not. Furthermore, personality traits could have an
influence on how synesthetes cope with these synesthetic affective reactions, so that
they do not have a great impact on their behavior. In Chapter 4 we tested MA and
found she has a high score in anxiety personality trait. Several authors have related
high anxiety traits with a lack of control (Lazarus, 1991; Mandler, 1984). This might
195
explain why she has problems controlling her synesthetic affective reactions, whereas
other synesthetes, which might have similar affective reactions, can control them
before they have an effect on performance. In this direction, in Chapter 4 we tested
PSV, who reported strong affective reactions associated to incongruently colored
stimuli but did not show a hint of a congruency effect either in the rating task or a
categorization task. PSV scored very low in the anxiety trait test. It does not seem to be
a coincidence that both factors correlate although, of course, in order to extract
conclusions, a sample of synesthetes, not single cases, would have to be tested.
Although further research is needed to test this hypothesis it seems a good candidate a
priori to explain some of the inconsistent patterns found.
To summarize, although more research is needed, the experiments reported in
this chapter suggest that only black is processed as a neutral color under certain
conditions. This might be due extensive and intensive practice with written material
printed in such color. As already noted by Dixon et al. (2006), meaning has an
important role in synesthesia and not only so in the process of experiencing one color
or another when perceiving a stimuli but, as shown here, also when categorizing a
stimulus as congruent or incongruent with the internal perception. Two factors seem to
have a crucial influence on these processes. Context influences how a color is
categorized (especially for black) and subjective interpretation of the perceived wave
length also seems to play a role (i.e. grey vs. perceived-as-white color). Some other
factors influencing the affective reaction might be personality traits or amount of
practice with the presented colors.

CHAPTER 6
SYNESTHETICALLY INDUCED AFFECTIVE
REACTIONS AND ATTITUDE FORMATION

Chapter 6. Affective Reactions and Attitude Formation ____________________________
199
ABSTRACT
Previous experiments have shown that affective reactions experienced as a
side effect of synesthesia are real and quite strong (at least in some synesthetes). Here
we wanted to test whether these affective reactions are strong enough as to condition
co-occurring events through classical conditioning. A visual search task was devised
where participants had to search for a target image shown in one of nine locations and
index the location number by hitting the appropriate key in the keyboard. Congruence
between the photism elicited by the number indexing the location and the color of the
target image was used to condition attitudes towards that target image. Results show
that when the conflicting dimensions of the stimuli are simultaneously shown on the
screen, later evaluation informs of a conditioning effect whereby those target images
associated to incongruently colored numbers and shown in that incongruent color are
perceived as less positive than images associated to congruently colored numbers and
shown in that color.
INTRODUCTION
Synesthetes usually report that the experience of synesthesia has an emotional
side, a sense of certitude that what is experienced is real (Cytowic, 1993;
Ramachandran & Hubbard, 2001b). In this sense, when synesthetes are exposed to
material incongruent with their synesthetic perception (i.e. words colored in the wrong
photism) they have emotional reactions that can vary from simple unpleasantness to
intense emotional reactions even accompanied of physiological reactions and escape
responses. Callejas, Acosta and Lupiez. (submitted, Chapter 3) have shown that this
affective reaction can be reproduced and studied in a laboratory setting, it happens
automatically, is difficult to ignore and influences behavior. We have also shown that
these results are not an idiosyncrasy of the synesthete we tested but are generalized to
other synesthetes from widely different backgrounds.
Going a step further, we next questioned whether this affective reaction would
be strong enough as to affect attitude formation about a stimulus that is otherwise
neutral or already has a positive or negative valence. Attitude formation has long been
studied by social psychology and it is thought to occur through classical conditioning
where the object for which an attitude will be formed (conditioned stimulus, CS) is
encountered repeated times paired with a stimulus that has a positive or negative
valence (unconditioned stimulus, US). Early research seemed to back up this idea
200
(Razran, 1938; Staats & Staats, 1958) and even proposed that such a mechanism
happens without awareness of the CS-US pairings, but further experiments cast doubts
about the lack of awareness in such conditioning process.
Recently Olson and Fazio (2001) carried out a study where they leaned on an
implicit learning paradigm to study whether such attitudes were formed implicitly. The
main differences between their paradigm and previously used ones was that they did
not ask participants to rate stimuli prior to the conditioning phase but obtained this
information form a different sample; they paired each CS with different US; both were
presented simultaneously with other events and they used a cover story and a
experimental task that ensured participants were paying attention while performing a
task different from the detection of covariance between paired stimuli. After the
conditioning phase in which participants had to detect an infrequent stimulus appearing
in a sequence of other stimuli, they asked participants to rate some of the stimuli
presented to make sure that affective reactions to some of them did not interfere with
the detection of the target stimuli. Among the rated stimuli were a couple of filler
images (US) that had been consistently paired with positive or negative attributes (CS).
A subsequent awareness test showed that participants were not aware of the CS-US
pairings shown in the conditioning phase.
Inspired by this experiment we sought to study whether the negative affective
reaction associated to a synesthetically incongruently colored stimulus would be strong
enough as to serve as a CS- and whether the positive affective reaction associated to a
congruently colored stimulus would be able to act as a CS+. An in our previous studies
(Callejas et al., submitted; Chapter 3) a congruently colored stimulus would be that
presented in the same color as the photism that our synesthete experiences for it and
an incongruently colored one would be the color-wheel opposite of the congruent color.
EXPERIMENT 1. IMPLICIT ATTITUDE FORMATION WITH IMPLICIT
SYNESTHETIC INCONGRUENCE AS THE UNCONDITIONED STIMULUS.
In this first experiment we tried to apply Olson and Fazios paradigm (2001) with
an US stimulus that was not physically present on the screen. We devised a task where
a 3 by 3 grid appeared on the screen and a target image would be presented on one of
the 9 slots created by the grid. Each slot was identified by a number that was
presented on its upper right corner. A colored target was presented on each trial on one
of the grid slots and participants were required to inform of the position of the target
201
stimulus by means of the keyboard. Since each number elicits a color, although
numbers were presented in black, a color was being perceived by our synesthete but
not by control participants. The color of the target image could be the same as the
photism elicited by the number identifying the slot in which it had been presented or a
different color. This congruency between target color and slots numbers elicited
photism was expected to elicit an affective reaction that would in turn act as the
Unconditioned Stimuli.
Since participants had to identify the slots number in order to give an accurate
answer about the position of the target stimulus, the number had to be processed. A
good amount of papers have shown that photisms are automatically elicited by the
inducer stimuli (Dixon, Smilek, Cudahy & Merikle, 2000; Mattingley, Rich, Yelland &
Bradshaw, 2001; Mills, Boteler & Oliver, 1999) and our own research with our
synesthete MA showed that she does experience strong automatic photisms for
numbers and letters (Lupiez & Callejas, 2006, Chapter 1). This ensured that MA
would experience the photism color associated to each number while control
participants would lack this color information. Although not necessary for the
experiment, we expected the color of the target image to be also processed since color
processing seems to be a preattentive feature (Treisman & Gelade, 1980). If target
color processing and photism elicitation occurred at the same time, both colors could
either be the same (when the target was presented in the same color as the photism
elicited by the number presented in the same slot) or different (when the target was
presented in a color different from that elicited by the number identifying the slot where
it had been shown). By manipulating the color of the target we could then create
incongruent trials and congruent trials. If this incongruence was experienced with the
sufficient strength, it could in turn condition the target stimulus which could be later
measured as an attitude towards that stimulus.
METHOD
Participants
MA and a set of 10 Psychology Undergraduates from the University of Granada
took part in this study. Their mean age was 20 and 2 of them were males.
202
A set of 14 target stimuli and 36 distracters were used in this experiment. Target
stimuli were Pokemon images. These images were chosen following Olson and Fazio
(2001) who successfully induced positive or negative attitude towards them by pairing
them with positive or negative words. Distracters were nonsense images designed as
to have the same perceptual features. These images are called freebles. Pokemon
figures were randomly divided into pairs and each pair was colored in a different color
(i.e. the colors evoked by the numbers 1-9 for MA). Once they had been colored we
proceeded to normalize them with a different sample of participants. Even though the
Pokemon figures chosen were not the most popular ones, participants were willing to
rate them on a pleasantness scale. One Pokemon from each pair was assigned to the
CS+and one to the CS-. Since some figures tended to be categorized as more positive
than others just based on their perceptual features, (i.e. face expression, round vs.
angular lines, etc) elements of each pair were assigned to one group or another so that
the mean rating for each group was equal. A representative subset of the distracter
freebles used as well as MAs colors for numbers 1-9 and the Pokemon images
associated to each number can be seen in Table I.
The screen was divided in a 3 by 3 matrix and was outlined resulting in a 3 by 3
table. Slots were numbered from bottom to top and from right to left being the bottom
right slot numbered 1 and the top left one numbered 9. This ordering was chosen so
that it did not resemble any ordering to which participants could be habituated to such
as the keyboard of a portable calculator (i.e. bottom to top, left to right) or the keyboard
of a cell phone (i.e. top to bottom, left to right) so that we could assure participants
were actively paying attention to the number in each slot in order to respond.
(Schneider, Eschman & Zuccolotto, 2002).
Two variables were manipulated in the experiment. Congruency referred to the
match-mismatch between the color of the Pokemon figure that the photism elicited by
the number of the slot in which the figure was presented. We could have congruent and
incongruent trials. Distracter had two levels: either one or three freebles could be
presented along with the target figure. Half the trials were presented with one distracter
203
and the other half with three distracters. Distracters could appear in any random
location other than the target one.
Table I. MAs colors for numbers, conditioned images and sample of distracters.
MAs
colors
CS+ CS- Distracters (freebles)
1

2

3

4

5

6

7

8

9

204
The experimental session was as follows. Participants were instructed that they
were participating in an experiment studying their ability to transpose one spatial
arrangement into another spatial arrangement. They would be shown stimuli in a 3 by 3
matrix and had to respond with the number keys of the alphanumerical keyboard
therefore having to deal with two spatial maps for numbers, the one on the screen and
the lineal one on the keyboard. They were instructed to look for the target image that
would be a Pokemon figure and inform as fast as possible about its location. They were
instructed to lay their fingers on the keys from one to four with the left hand and six to
nine with the right hand and respond to five with either index finger. Speed and
accuracy were requested.
The locations matrix was presented throughout the whole block. Each trial
began with the locations matrix presented alone. After one second, the images
appeared in some of the slots and were present until a response was given. After the
response the next trial began. Since two numbers elicited a red photism (2 and 5) and
two numbers elicited a yellow photism (1 and 7), the red CS+image as well s the
yellow CS+image were shown once in each of the two locations eliciting the color
figure. To equal the number of times each Pokemon image was shown, the other five
Pokemon images were shown twice in their corresponding location. The way we
proceeded for the CS- Pokemon was the same. Those CS- Pokemon images shown in
yellow and red were presented only once in each one of their two possible locations
while the rest of the Pokemon images were presented twice in their location. Therefore
in each block we presented each one of the 7 CS+Pokemon images twice and each
one of the 7 CS- Pokemon images twice. This resulted in 56 trials per block (7
Pokemon images x 2 presentations x 2 Congruency x 2 Distracter)
After participants had run 10 practice trials and five blocks of 56 experimental
trials each, they turned to the second part of the experiment. Here they were told that
experimenters would like to use those same stimuli for other emotion experiments and
we would like them to rate the images in a 7 point scale with -3 being very negative and
+3 very positive. Four blocks of trials were presented and in each block 28 images
were shown one by one: the fourteen Pokemon images as well as 14 of the 36 freebles
used. On each trial a reproduction of the responding scale was presented along with
the fixation point for 500ms. After this, the fixation point was replaced by the image to
be rated. Once a rating was emitted, the next trial began.
Last, participants were told that the images had always appeared in the same
position and experimenters were interested in knowing whether they could remember
205
the position. They were asked to inform of the position of each figure by pressing the
corresponding number key. Again, after a 500ms fixation point the image was
presented and kept on the screen until participants hit one of the number keys. The 14
Pokemon figures were presented as well as a subset of 14 freebles included as fillers
so that participants would think that freebles were also always presented in the same
location.
Localization task
Since we had to compare MA to the group of ten controls, we used again items
(the 14 Pokemon images) instead of participants as the random factor. Therefore
reaction times were averaged across participants and blocks for the control group and
across blocks for MA. Trials were the correct location was not identified accounted for
8.3% for the control group and 1% for MA and were excluded from further analysis. A
2SD cut off was applied to eliminate extreme values. Since Pokemon images were the
random factor, congruency between the images color and the photism elicited by the
location number was a between items factor while number of distracters and group
were within item factors. We therefore carried out a mixed 2 (Congruency) x 2
(Distracter) x 2 (Group) ANOVA that yielded the following results (see Table II).
The analysis showed that MA was about 70ms faster to respond than the
control group (F
(1, 12)
=8.11, p=.015). Also, pointing out the location of the Pokemon
image was faster when only one distracter was present on the screen as opposed to
three (F
(1, 12)
=25.79, p=.001). However Distracter did not interact with Group or
Congruency thus showing that MA performed the task as control participants did.
Table II. Experiment 1. Mean RT and error rate for Controls and MA as a function of
Congruence and Distracter number.
Position 1 Distracter 3 Distracters 1 Distracter 3 Distracters
Controls
1291.62
(7.3%)
1330.33
(10.0%)

1294.10
(8.6%)
1366.04
(8.7%)
MA
1228.74
(0.0%)
1262.21
(0.0%)

1219.21
(0.0%)
1282.09
(3.6%)
206

Regarding error rates, MA was more accurate than controls since she was
errorless when one distracter was present and 1.79% errors were made when three
distracters were present. Control participants on the other hand were less accurate
although they did also show a slightly better performance when only one distracter was
present (7.96% vs. 9.36%). In regards to congruency, all errors that MA committed
were when three distracters were present and the color of the Pokemon image was
incongruent with the photism elicited by the slot number. No statistical analysis was
carried out.
Conditioning effect
Next we analyzed the conditioning effect. Ratings for each participant were
analyzed separately. A paired samples t-test was carried out to compare each
participants ratings for the CS- items and CS+items. The rating for each time each
image was presented were taken as independent measures. Since each one of the 7
CS- Pokemon and the 7 CS+were presented four times, 28 observations per CS group
were introduce into the test. Ratings for the filler freebles were not analyzed since they
had not been consistently paired with any number. Table III shows the mean rating for
CS- and CS+as well as the results for the t-tests for the ten control participants and
MA. As can be seen in the table, control participants did not rate negatively conditioned
Pokemon images as more negative than positively conditioned ones. Only two
participants showed a significant difference and it was in the direction opposite to the
one expected. This result confirms that both groups of stimuli had been correctly
configured as to give rise to equal evaluation when no other factor was involved. Last,
we can also see that MA did not show a conditioning effect either. Her mean rating for
CS- was 1.04 and for CS+it was 1.18.
Table III. Experiment 1. Valence ratings for CS- and CS+ and percentage of positions
remembered for each control and for MA.
Mean rating Percentage position remembered
Subject CS- CS+ T-test value CS- CS+
1 -0.57 -0.36 T(27)= -0.57; p=.577 28.6% 28.6%
2 -0.11 -0.25 T(27)= 0.21; p=.834 42.9% 42.9%
3 1.61 0.46 T(27)= 2.41; p=.023 * 57.1% 28.6%
4 0.96 0.21 T(27)= 1.38; p=.179 14.3% 42.9%
5 1.18 1.75 T(27)= -0.97; p=.341 42.9% 28.6%
6 0.50 0.29 T(27)= 0.39; p=.709 14.3% 14.3%
7 -0.18 0.11
T(27)= -0.42; p=.678
0.0% 14.3%
207
8 -0.96 0.11 T(27)= -1.90; p=.068 71.4% 14.3%
9 0.14 -1.36 T(27)= 3.45; p=.002 * 14.3% 14.3%
10 2.61 2.25 T(27)= 1.67; p=.106 28.6% 14.3%
MA 1.04 1.18 T(27)= -0.48; p=.636 28.6% 14.3%
Controls
Mean
0.51 0.32 T(27)= 0.59; p=.560 31.4% 24.3%

Location recall
Table III also shows the percentage of times the location of the Pokemon image
was correctly recalled. It can be seen that most participants did not show a good
memory for the position in which the stimuli had appeared. MA was among those
showing the least recall.
Data from this experiment shows that our procedure was not successful
inducing attitudes towards otherwise neutral stimuli. Reaction time and accuracy data
shows that participants were performing the task correctly and therefore paying
attention to the number identifying the location where the target Pokemon appeared.
The conditioning effect analysis shows however that MA did not rate positively
conditioned Pokemon images as significantly more positive than negatively conditioned
ones. As expected, control participants did not show differences either since numbers
did not have the extra color dimension that was expected to interact with the Pokemon
color. Although 2 control participants did show a difference between both groups, it was
in the opposite direction as that expected (i.e. they rated negatively conditioned images
as more positive than positively conditioned ones).
One explanation for this lack of an effect could be that images can not be
conditioned with such an indirect method. However, another reason could be that we
were not sufficiently promoting the processing of the number in order for it to elicit a
photism and interact with the color of the Pokemon image. The fact that slots had fit
numbers could help participants learn a spatial location to finger association that would
automatically occur after a certain number of trials without activating the number
concept associated to that slot. If the slot to response finger mapping was strong
enough, numbers would not be needed in order to respond accurately and therefore
would not be attended thus not eliciting a photism. The fact that MA showed such a
high accuracy in the conditioning section of the experiment and she was also faster
than controls suggest that she was in fact responding based on a location-to-finger
208
mapping basis and not on a number-to-keyboard-number mapping fashion. Also, the
fact that a speeded response was requested could have made it less probable for the
number to elicit a strong photism that would subsequently interact with the color of the
target image. Previous studies have shown that allocation or withdrawal of attention to
the inducer stimulus can modulate the degree to which the photism influences
performance (Rich & Mattingley, 2005; Sagiv & Robertson, 2005). Therefore it could be
that the photism for each number was not strongly experienced and thus had not
interacted with the Pokemon color.
In order to address all these issues we conducted a second experiment. To
minimize possible effects of this experiment on the subsequent one, a 3 year lapse was
introduced between experiments.
EXPERIMENT 2a. IMPLICIT ATTITUDE FORMATION WITH EXPLICIT
SYNESTHETIC INCONGRUENCE AS UNCONDITIONED STIMULUS
In order to address the possible facts influencing the null effect found in
Experiment 1 we designed this second experiment. We first changed the way slots
were numbered. Three different numbering systems were used and in each trial a
different one could be present. This way we ensured participants would actively pay
attention to the number presented in the slot in order to give an accurate response.
Also, we presented the slot numbers in different colors depending on the type of trials.
In those trials were a CS- was presented all the slot numbers were colored
incongruently with the elicited photism. The colors used were the same ones assigned
for the CS- Pokemon. Likewise, in those trials were a CS+ was presented all the
numbers in the slots were colored congruently with the photism they elicited. With
these two manipulations we intended to avoid the issues that could be at the basis of
our null results in Experiment 1.
METHOD
Participants
MA and a group of 10 Psychology Undergraduates run this experiment. Control
participants were different from those running Experiment 1. MA run this experiment
three years after the first one to assure she did not remember the specifics such as the
209
fact that each Pokemon image was always presented in the same location. Control
participants mean age was 21 and 2 of them were males.
The very same target images and distracter images used in Experiment 1 were
used here. Three different numbering systems were used to identify each slot. The first
one was as that of Experiment 1 (i.e. bottom to top-right to left). The second system
was also right-to-left based but the first three numbers were on the top, the second
three ones on the bottom and the last three numbers in the middle row (i.e. 1 was on
the right in the top row and 9 was on the left in the middle row). The last system was
left-to-right based with the first row in the middle, the second one on the top and the
last one on the bottom (i.e. 1 was on the left in the middle row and 9 was on the right in
the bottom row). Everything else was as in Experiment 1.
The same two variables, Congruency and Distracter, were manipulated. as in
Experiment 1. Since the task was somewhat more difficult now due to the random
numbering of the slots, we introduced a feedback sound after response for those trials
were a mistake was made. The numbering systems appeared randomly across the
whole experiment.
Also, since the number of conditions was larger now (56 trials x 3 different
numbering systems) we run 6 experimental blocks with 56 trials each so that there
were two repetitions of each trial type.
Again after the localization task participants rated the valence of each stimuli
with the excuse that they would be used for another emotion experiment and they last
informed of the position in which each image had appeared.
Localization task
Again we averaged RTs across blocks and participants for the control group and
across blocks for MA. Control participants error rate was 2.44% while MAs was 1.2%.
A 2SD cut off was used to eliminate extreme values. Table IV shows mean localization
RTs and error rates per group and condition.
210
A mixed 2 (congruency) x 2 (Distracter) x 2 (Group) ANOVA was carried out on
the filtered data. We again found that MA was about 45ms faster than the control group
to inform about the location of the Pokemon image (F
(1, 12)
=7.29, p=.019). We also
found that the target image was identified faster when only one distracter was present
on the screen (F
(1, 12)
=40.77, p=.001). Again we found no other significant effect in the
localization data.
Table IV. Experiment 2. Mean RT and error rate for Controls and MA as a function of
Congruence and Distracter number.
Position 1 Distracter 3 Distracters 1 Distracter 3 Distracters
Controls
1230.41
(2.6%)
1339.00
(2.1%)

1217.94
(2.3%)
1279.72
(2.7%)
Session 1
1208.05
(1.2%)
1253.36
(1.2%)
1180.99
(1.2%)
1241.27
(1.2%)
MA
Session 2
1247.88
(0.0%)
1287.63
(2.4%)

1214.81
(0.0%)
1266.56
(0.0%)

In regard to the accuracy measures controls as well as MA committed the same
percentage of errors independent of number of distracters or congruency condition (for
controls 2.5% errors for one distracter and 2.5% errors for three distracters; for MA,
1.2% errors for one distracter and the same percentage for three distracters). No
statistical analysis was carried out.
Conditioning effect
Again we analyzed ratings for each participant independently by means of t-
tests comparing CS- with CS+. Table V shows the mean rating for each participant as
well as the result of the t-test. Again, controls did not show a conditioning effect. Only
participant 4 and participant 6 showed a significant difference between both ratings and
only in the case of participant 6 did it go in the expected direction. It could be argued
that since participant 6 showed an effect, then if we also found it in MA there was no
reason to rely on emotional reactions associated to inconsistent stimuli to account for
the results. We could simply argue that the perceptual inconsistency between the color
in which the number was presented and the color of the target image that had to be
responded to it made the task more difficult and that difficulty was in turn what made
those Pokemon images be liked less. However, the fact that only one out of the 10
control participants showed the pattern expected for MA suggests a different story. As
211
mentioned in the first experiment, the images used did have a valence associated to
them and we spitted them so that both groups would have similar ratings. It could well
be that the a priori valence that participant 6 was not even maybe due to greater
exposure to Pokemon cartoons and that explained the pattern found in the a posteriori
valence rating.
Table V. Experiment 2. Valence ratings for CS- and CS+ and percentage of positions
remembered for each control and for MA.
Mean rating Percentage position remembered
Subject CS- CS+ T-test value CS- CS+
1 1.61 0.36 T(27)= 1.80; p=.083 14.3% 0.0%
2 0.46 0.32 T(27)= 0.19; p=.849 14.3% 28.6%
3 0.64 0.93 T(27)= -0.75; p=.460 14.3% 14.3%
4 1.79 0.82 T(27)= 2.31; p=.028 * 14.3% 0.0%
5 0.64 0.79 T(27)= -0.21; p=.833 0.0% 14.3%
6 1.79 2.07 T(27)= -2.12; p=.043 * 71.4% 28.6%
7 0.86 0.89 T(27)= -0.07; p=.941 0.0% 71.4%
8 -0.75 -0.61 T(27)= -0.11; p=.846 42.9% 42.9%
9 0.79 0.50 T(27)= 0.71; p=.486 14.3% 28.6%
10 0.68 0.29 T(27)= 0.54; p=.596 28.6% 42.9%
MA 1.57 2.00 T(27)= -1.34; p=.190 0.0% 100.0%
Controls
Mean
0.85 0.64 T(27)= 0.64; p=.531 21.4% 27.2%

As can be seen, MA did evaluate CS+ as more positive than CS- but this
difference was so small that it did not reach statistical significance. We will come to this
result later.
Location recall
As shown in Table V, on average, control participants recalled about one fourth
of the locations (27.14% for CS+and 21.43% for CS-). MA on the contrary recalled the
exact location of all the congruent stimuli (CS+: 100%) and none of the locations where
the incongruently colored Pokemon had appeared (CS-: 0%). This is an interesting
result that points out two different things. One, MA was influenced by the congruency
between the color of each number and the color in which the Pokemon was presented
as opposed to Experiment 1 where the number was not colored. Second, this result
resembles that found by Smilek, Dixon, Cudahy and Merikle (2002) where they showed
that congruently colored stimuli were remembered much better than incongruently
212
colored ones being the congruency between the presented color and the photism
elicited by the stimuli presented.
This new experiment went in the same direction as Experiment 1. Although MA
was now more conscious of the location in which the congruently colored Pokemon
images appeared, she was not so for the incongruent ones. Most important, although
she did show a small difference between CS- and CS+, it did not reach significance. It
seemed then that the procedure we devised either did not elicit the affective reactions
associated to perceiving an incongruently colored stimulus or they did but such an
affective reaction was not strong enough as to serve the function of an US. However, a
closer look to MAs recall pattern revealed a very interesting fact. When checking her
actual responses for the location of the colors we realized she was just guessing and
not actually remembering their location. MA chose for each Pokemon image the
location of the number with a photism that matched the color of the Pokemon.
Therefore this strategy yielded 100% accuracy for the CS+Pokemon images and 0%
for the CS- Pokemon images. The fact that she did not guess for the CS- images
makes us think that either she did not remember the location of any of them and just
used the color as a link to guess a probable location or that she did realize that one
Pokemon from each color pair was in the location congruent with such color but the
other one was not but she was not able to discern which one was so she just guessed
for both in hopes that at least one from each pair would be right.
EXPERIMENT 2b. SECOND SESSION
Since this closer look to the data suggested she was not aware of the
underlying motives for this experiment and the consistent pattern in which the images
were presented, we decided to run this experiment once more to check if the lack of
significance in the conditioning effect analysis was due to a high variability of the data.
METHOD
Participants
Only MA run this second session.
The same stimuli used in Experiment 2a was used here.
213
Three weeks after MA had run Experiment 2a we called her up again and asked
her to run it again. Since she is used to coming back for second sessions so that we
obtain a more stable estimate of her performance, she was not surprised to run the
experiment again with that excuse.
This time, in order to check whether the rating pattern found in Experiment 2a
was still there three weeks later, we started by asking her to rate the valence of each
one of the images again. In order to obtain a more stable estimate, each image
appeared three times in each of the four blocks. Therefore 84 samples from each
group (i.e. CS- vs. CS+) were compared.
Right after a short break we run the whole experiment again. Here the Pokemon
evaluation phase was as the above mentioned one with three presentations of each
image in each block.
On this previous evaluation phase we found that she rated one group of
Pokemon images as positive as the other group (1.20 for CS- and 1.20 for CS+).
Therefore the little difference found in Experiment 2a was extinguished.
The results from this second session (Experiment 2b) are reported below and
can be seen in Table IV. The mixed 2 (Congruence) x 2 (Distracter) ANOVA informed
that responding when only one distracter was present was faster than when two
distracters were present (F
(1, 12)
=5.73, p=.034). Congruency did not affect reaction
times nor did it interact with Distracter.
Regarding the error rates, MA was again very accurate and she only deviated
from perfect performance when congruent images were shown with 3 distracters
(2.38% errors).
If we turn to the recall data, we found the exact same pattern as in the first
session of Experiment 2. MA correctly located each of the CS+Pokemon images and
did not get one single location right. Again, she assigned the location as a function of
their color. Even more interesting, since two numbers elicited yellow and two numbers
elicited red, she assigned each one of the yellow Pokemon images to one of those two
numbers and each one of the two red Pokemon images to one of those two red elicitor
214
numbers. Again, it seemed as though MA was not aware that half of the Pokemon were
shown in a location different from that where the color of the image matched the
photism elicited by the identifying number.
Table VI. Experiment 2b. Valence ratings for CS- and CS+ before and after the
conditioning phase and percentage of positions remembered after the conditioning
phase for MA.
Mean rating
Percentage position
remembered
Measurement
moment
CS- CS+ T-test value CS- CS+
Pre
conditioning
1.20 1.20 T(83)= 0.00; p=1.0
Post
conditioning
-0.31 1.49 T(83)= -5.63; p=.001 0.0% 100.0%

Last, we checked her valence ratings for those same images she had rated 30
minutes before as equally pleasant (1.20 for CS+and CS-). We now found that CS+
obtained a mean rating of 1.49 and CS- obtained a mean rating of -0.31. A t-test
confirmed that this difference was statistically significant (T(83)= -5.63; p=.001).
Therefore it seems as though a reliable conditioning effect was found and MA rated
those Pokemon shown in a color different from that of the number signaling its position
as more negative than those Pokemon paired with a number in its right color.
GENERAL DISCUSSION
With this study we wanted to further test the strength of the affective reactions
reported by many synesthetes when encountering a stimulus in a color different from
that elicited by such stimulus for a particular synesthete. Synesthetes reports of these
affective reactions have been gathered by a few researchers (Cytowic, 1993;
Ramachandran & Hubbard, 2001b) and found to be automatic, strong and difficult to
ignore by Callejas et al. (submitted, Chapter 3). Here we wanted to test whether the
strength of such affective reactions would lead to the conditioning of neutral stimuli.
Using a novel paradigm introduced by Olson and Fazio (2001) we tried to
condition a set of Pokemon images (CSs) using the affective reaction elicited by the
mismatch between photism elicited by the number identifying the location of the image
and the color of the image itself as USs. Results showed that this manipulation was not
appropriate to elicit such an evaluative conditioning. In a second experiment we tried to
215
more clearly state the stimulus dimension that would give rise to the affective reaction.
In order to do so we used three different numbering systems to ensure participants
engaged in a number to number mapping when responding and not in a location to key
mapping. We also presented the numbers identifying the positions either in their
congruent color or in an incongruent color so that both number color and target image
color would be present simultaneously. This time, a small effect was found in MAs
attitudes toward positively and negatively conditioned stimuli, although this difference
was far from statistical significance. However, results after a second session showed
that she did acquire an attitude towards the CS+that was more positive than that
shown towards the US-.
On the other hand, an important conclusion can be extracted from the results of
the location recall analysis. Note that all stimuli (both CS+and CS- of each color) were
recalled as having been presented in the location identified by the number eliciting a
photism that matched the image color. This is why memory was 100% correct for CS+
stimuli, which had been in fact presented in the position corresponding to its color, and
0% correct for CS-, which never were presented in the corresponding position.
Therefore, the fact that location recall was perfectly driven by the color in which the
Pokemon images were presented led us to think that MA was not aware of the
consistent association of half the images to locations eliciting a different photism than
that of the image. Also, her repeated comments regarding this experiment being one of
the experiments she had done that she enjoyed the most (as opposed to her reports
after Stroop or Valence Categorization experiments were she complained of many
items being colored wrong) backed up our beliefs that she was not conscious of the
Congruency manipulation.
We therefore conclude that the affective reaction triggered by an incongruently
colored stimulus acted as an unconditioned stimulus and conditioned an attitude
towards an otherwise neutral stimulus.
The fact that the effect was not found in the first session could be due to an
insufficient number of conditioning trials for each Pokemon image. Although CS+
images were shown 168 times and CS- images another 168 times total, each Pokemon
was presented only 24 times paired with its corresponding colored number. It could well
be that these were not enough conditioning trials to evoke a statistically reliable effect.
Having previously found that MA rated those same Pokemon images equally
independent of their CS group assured that she did not have an a priori differential
216
preference for any of them and thus the results found in the second session of
Experiment 2 can not be attributed to personal preferences for certain Pokemon
images.
These results have wide implications in the understanding of how synesthesia
influences synesthetes and their way to go about in a normal day basis. If a couple of
hundred trials showing an image paired with a certain incongruence results in a
negative attitude towards that image, a lifetime of encounters with incongruent stimuli
paired with different objects could be determining their attitudes towards situations or
objects such as a milk bottle for which the TV commercial shows the brand name next
to the word milk in an incongruent color. Research in the field of marketing and
advertising has been interested in classical conditioning as a form of attitude formation
(Allen & J aniszewski, 1989; Kim, Allen & Kardes, 1996) and it is now the time to see
how those findings of non-synesthetes relate to people experiencing different types of
synesthesia.

CHAPTER 7
ARE SYNESTHETES THE END OF A CONTINUOM?
COLOR EMOTION RELATIONS IN NON-
SYNESTHETES

Chapter 7. Color-Emotion Reactions in Non-Synesthetes __________________________
219
ABSTRACT
It has been shown that synesthetes experience affective reactions as a
consequence of interacting with a stimulus presented in a color different from the one
they experience due to their synesthesia. Color congruence effects have also been
found in non-synesthetes as a function of the color in which words are presented but
no study has tested the factors underlying this unexpected and interesting phenomena.
Here we test three variables that could modulate emotional categorization of words.
We presented a group of non-synesthetes with a set of emotional words to be
categorized as positive or negative according to their meaning. Words were shown in
the four basic colors (i.e. yellow, blue, red and green), and either above or below the
fixation point. Color preference and subjective color lightness were coded subsequently
and data was analyzed as a function of these two variables together with the position.
The three variables had an influence on the emotional categorization responses.
Results show that participants color preference influences their categorization ratings in
the same manner as color-photism match-mismatch does in synesthetes.
INTRODUCTION
Affective reactions associated to the perception of incongruently or congruently
colored stimuli have long been reported by synesthetes and captured in experimental
papers (Cytowic, 1993, 1995; Ramachandran & Hubbard, 2001b). Callejas, Acosta &
Lupiez (submitted, Chapter 1) studied this issue experimentally. In a first experiment
they asked MA, a previously studied female synesthete (Lupiez & Callejas, 2006), to
rate the emotional valence of different words. Words were either congruently or
incongruently colored, according to MAs photism for each word. Results clearly
showed that the ratings of all kind of words were affected by their coloring, although
color was completely irrelevant to the task. MA rated words as more negative when
incongruently colored and more positive when congruently colored.
In order to assess the automaticity of this effect of color-photism congruence on
emotional valence rating MA was asked to categorize words as either positive or
negative according to their meaning. In order to make sure that MA followed
instructions to focus attention on the meaning of the words, while ignoring the color,
authors put great emphasis on accuracy and provided feedback on wrong
categorization responses. In spite of this, results again showed that coloring affected
emotional categorization. The congruency effect typically seen in synesthetes when
________________________________SECTION 2. SYNESTHESIA AND EMOTIONS
220
they have to respond to a stimulus colored inconsistently with their synesthetic
perceptions (i.e., faster responses to congruently colored than to incongruently colored
stimuli) was only observed under certain conditions whilst it was eliminated or even
reversed under certain other conditions. In the case of semantically negative words,
MA was faster to categorize them as negative when they were incongruently colored
than when they were colored congruently with their photism.
The authors proposed that a double assessment being carried out by
synesthetes could explain the observed pattern of results. Synesthetes assess the
semantic valence of the word but they also assess the congruence of the word with
their subjective perception. If a positive word is colored right (i.e. with the color of their
photism for that word), the outcome of both assessments is of the same sign. Similarly,
if a negative word is presented in the wrong color, the outcome of both assessments
is again of the same sign (in this case both outcomes would be negative). Alternatively,
when a positive word is presented in the wrong color or a negative word is presented
in the right color, the outcome of both evaluation processes is of a different sign. This
inconsistency is ultimately what drives their performance increasing their RT and
decreasing accuracy in the task. Further studies replicated this finding with another
group of synesthetes (Chapter 4).
It is interesting to note however that in all experiments a congruency effect was
also found for the control participants that do not experience synesthesia or explicitly
reported experiencing any affective reaction when faced with words in a particular
color. This congruency effect, although of a much smaller magnitude, was randomly
found either in RT or accuracy measures and in the same line as the effect found in
synesthetes or in the opposite direction. Since it did not follow any understandable
pattern, it was discarded as an unknown effect of the colors used.
Previous work have posited the possibility that synesthesia shares the same
neural mechanisms as normal multi-sensory perception (Gevers, Reynvoet & Fias,
2003; Rich, Bradshaw & Mattingley, 2005; Sagiv & Ward, in press; Simner et al, in
press; Ward, 2004; Ward, Huckstep & Tsakanikos, 2006, to cite some) and have found
some similar trends between synesthetically experienced colors and colors that non-
synesthetes tend to associate with the same stimulus. In this sense, synesthetes would
be not qualitatively different from non-synesthetes, but the extreme pole of a
continuum.
221
Therefore, a possible explanation for the random effects that we found in the
previous experiments with control participants could be that some underlying trends of
color-emotion associations found in control individuals (non-synesthetes) would be
influencing their semantic valence categorization of words. Supporting this explanation,
early studies have associated less saturated colors with negative emotions and lighter
and more saturated colors with positive emotions (Collins, 1929; Cutsforth, 1925;
Hemphill, 1996) although some other researchers suggest that more liked colors are
associated with more positive emotions and less liked colors with more negative
emotions (Terwogt & Hoeksma, 1995). It could then be that one or both of these factors
would be playing a role in our previous experiments and thus introducing noise in the
overall results of the control group. Since reanalyzing this data on the basis of the fore
mentioned factors was quite unattainable (for one thing, we could not go back and ask
participants which color they liked more, and for another thing, congruent and
incongruent colors were tailored for each synesthete), we decided to carry out a new
experiment in which we systematically measured these two dimensions and tested how
they affected the categorization responses.
Therefore, our aim in this study is to test whether the effects found in those
previous works could be explained on the basis of subjective color preferences or
subjective perceived lightness of the colors used. In order to do so we asked
participants, as in previous experiments, to categorize emotional words as either
negative or positive. All words were presented in different colors and, after the
categorization task, participants were asked to rate these colors according to their
subjective preference and their subjective lightness perception. These 2 factors were
subsequently used to analyze their performance on the categorization task.
EXPERIMENT 1. VALENCE CATEGORIZATION AS A FUNCTION OF COLOR
PREFERENCE AND SUBJECTIVE COLOR LIGHTNESS
In this first experiment we tested the relative influence of the four primary colors
on a valence categorization task. Previous studies had suggested a link between color
and emotion (Adams & Osgood, 1973; Kreitler & Kreitler, 1972; Terwogt & Hoeksma,
1995). We expected that if participants had clear preferences for a particular color and
clear dislike for a particular color, their performance in the valence categorization task
would be influenced by those colors and a pattern of results similar to that found in
synesthetes would emerge. Positive words would be easily categorized when colored
with the preferred color and negative words would be easily categorized when colored
222
with the disliked color. Alternatively, if colors are associated to either positive or
negative emotions according to their lightness (lighter colors associated to positive
emotions and darker colors to negative emotions), the positive words would be easily
categorized when colored with light colors, whereas negative words would be easily
categorized when colored with darks colors.
Finally, we were also interested in studying the emotion-space metaphor by
which emotions are associated to the up-down axis. It has been proposed that all
concepts are structured through metaphoric mappings, being representative of one of
the proposed primary metaphors the mapping of negative emotions down and positive
emotions up (Lakoff & J ohnson, 1980). Provided evidence for this metaphor, Meier and
Robinson (2004) showed that positive words presented on the top half of the screen
were categorized faster than when presented in the bottom half and conversely, for
negative words, they were categorized faster when presented on the bottom half. In
order to test this hypothesis in our study we randomly presented colored words above
or below the fixation point. Our results would support the psychological reality of the
positive emotion is up-negative emotion is down metaphor, if categorization responses
for positive words were faster above fixation, whereas categorization responses for
negative words were faster below fixation.
METHOD
Participants
A set of 39 Psychology undergraduate students participated in this experiment
in exchange for course credits. Seven of them were males and participants mean age
was 21.
A set of 36 emotional words was used here. Half of them were negative and the
other half were positive. These words were the same used in Experiment 2a of Callejas
et al. (submitted, Chapter 3). Words subtended 4.7 to 9.9 horizontally and 1
vertically. Each word was presented in one of four colors: yellow (255,255,0), blue
(0,0,200), green (0,225,0) and red (255,0,0). Words were presented on a neutral grey
background (125,125,125)
223
(Schneider, Eschman & Zuccolotto, 2002)
Words were presented in one of two different locations (above or below fixation
point) and in one of four different colors. Therefore, three a priori variables were coded:
word Position (up-5.3 above fixation point-, down-5.3 below fixation point-), word
Color (yellow, blue, red and green) and word Valence (positive, negative).
Participants run 8 blocks of 72 trials each (a total of 576 trials: two trials per
word, color and position). Conditions were randomly selected. Ten practice trials
randomly chosen from the 72 possible conditions were run prior to the experimental
blocks.
After these 8 experimental blocks, participants were asked about their color
preferences for the colors presented and their subjective lightness. In order to do so,
pairs of color patches were presented on the screen and participants had to choose
one color from each pair. Each color was paired with the other three colors totaling 6
different combinations (12 if we take into account the two positions in which the pair of
colors could be presented one color of the pair either to the left or to the right of the
other-). Each of the 12 combinations was presented on 3 different occasions in a
random order to obtain a more stable judgment. In the first assessment block
participants were asked to choose the color of the pair they liked the most. In the
second of these assessment blocks participants were asked to choose the color they
thought was the lightest as opposed to the darkest one.
Two new variables were coded once the experiment was run to take into
account participants responses in these last two blocks. These variables were called
color Preference (most liked vs. most disliked) and color Lightness (lightest vs.
darkest). The procedure used to code these variables is explained in the results
section.
At the beginning of each trial a black fixation point appeared on a grey
background for 500ms. It was then followed by the target word that was kept on the
screen until subjects response. Participants task was to categorize words as positive
224
or negative according to their meaning. Auditory feedback was given if the word was
not categorized correctly. After response (or feedback sound) the next trial began.
For the last two blocks, after fixation, two colored rectangles appeared at either
side of the fixation point. Participants had to choose the preferred one or the lightest
one (depending on the block) by pressing the key that was aligned with the color in the
screen (left key if they chose the color of the left rectangle and right key if they chose
the color of the right rectangle). As mentioned above, each color pair was presented in
both positions so that half the time one color of the pair appeared on the left and the
other half on the right. Rectangles were presented on the screen until response, after
which the next trial began.
Trials were words had been incorrectly categorized (5.6%) were eliminated from
the RT analysis. A 2SD cut off was set to eliminate outliers. Two different analysis were
carried out, one taking items (I, i.e., words) as random factor (as in Experiment 2a of
Callejas et al., submitted, Chapter 3) and another taking participants (P) as random
factor
1
. Response times for each word and coloring condition were averaged across
blocks and participants. A mixed ANOVA was carried out on the a priori set of variables
with 2 (Valence), 2 (Location) and 4 (Color) as factors. Table I shows the mean reaction
times and error rates for the item analysis.
Table I. Experiment 1. Mean reaction times and error rates (in parenthesis) per
experimental condition for the general analysis.

1
Graphs and Tables shown were elaborated with the means for the item analysis.
Valence
Positive Negative
Position Yellow Blue Red Green Yellow Blue Red Green
Top
710.60
(7.1%)
715.66
(6.6%)
748.93
(8.4%)
701.59
(6.6%)

727.89
(10.0%)
722.83
(9.7%)
727.71
(8.3%)
728.45
(13.2%)
Down
724.71
(7.5%)
725.24
(6.5%)
760.24
(9.5%)
727.80
(6.2%)

732.42
(10.0%)
724.76
(10.6%)
727.40
(8.5%)
732.38
(12.5%)
225

660
685
710
735
760
Negative Positive
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)
Top Down

Graph 1. Experiment 1. Reaction times as a function of valence and location. Error bars
Words presented on the upper half of the screen were responded to faster than
those presented on the bottom half (I: F
(1, 34)
=10.70, p=.002; P: F
(1, 38)
=6.91, p=.012).
This location effect was modulated by Valence (I: F
(1, 34)
=5.50, p=.025; P: F
(1, 38)
=6.47,
p=.015). As shown in Graph 1, when words were negative, the location in which they
appeared did not influence response times (I: F<1; P: F<1). However, when positive
words were presented on the upper part they were responded to faster than when
presented in the lower part (I: F
(1, 34)
=15.77, p=.001; P: F
(1, 38)
=18.63, p=.001).
650
700
750
800
Negat ive Pos it ive Negat ive Pos it ive
Up Down
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)
Yellow Blue Red Green

Graph 2. Experiment 1. Reaction times as a function of color, valence and location. Error
bars denote standard error of the mean.
The Color in which the word was presented also influenced categorization times
(I: F
(3, 102)
=20.38, p=.001; P: F
(3, 114)
=20.97, p=.001). Red was slower than any other
226
color (I: all ps<.001; P: all ps<.05) and all other colors did not differ (I: all ps>.9; P: all
ps<.9). This influence was modulated by Valence (I: F
(3, 102)
=23.19, p=.001; P: F
(3,
114)
=13.64, p=.001). As can be seen in Graph 2, red positive words were slower than all
other positive words (I: all ps<.001; P: all ps<.05).
This result clearly shows that red does not seem to be associated with positive
emotions, since positive red colored words where slower than any other word type
irrespective of subjective color Preference or Lightness. Previous work carried out in
the field of anthropology (DAndrade & Egan, 1974) showed that across cultures, terms
with negative connotations were paired with purplish and yellow-red hues while terms
with positive connotation were more consistently paired with blue-green colors. In
respect to location, we found that it does in fact play a role in the conceptualization of
emotions although its specific influence is yet not clear-cut.
In order to discern whether these effects were modulated by participants
preferences in color and their subjective perception of lightness, we carried out another
analysis with the aforementioned a posteriori coded variables Color Preference and
Color Lightness. We first coded the number of votes that each participant assigned to
each color in the Preference J udgment and in the Lightness J udgment. Graph 3 shows
the results for those ballots. A subsequent ANOVA was carried out in order to check
which of those differences were statistically significant. The analysis for Preference
showed that green was the most preferred color and red the most disliked color. Blue
and yellow obtained intermediate values. A post hoc test showed that only red was
different from the other three colors (all ps<.5). For the Lightness ballot we found that
yellow and green were the lightest colors (no difference among them, p<.07) and blue
was the darkest one (all ps<.02).
By looking at Graph 3 we cannot discern whether the slowest RT for positive
words presented in red is due to red being the most disliked color or to red being one of
the darkest colors. Therefore, in order to distinguish between these two possibilities we
recoded the color variable as color Preference and color Lightness. In order to code for
Preference we assigned a value from 1 to 4 to each color for each participant
according to their Preference votes. The four-value was assigned to the most voted
color and the one-value was assigned to the least voted color. We then recoded the
different conditions as a function of the assigned values. This way, the subsequent
analyses did not take into consideration the specific color that was presented but the
subjects rating for that color and so the most preferred color could be yellow for one
227
subject and blue for another. The coding of the Lightness variable followed the same
logic but now using the Lightness voting in order to assign values to colors.
0
2
4
6
8
10
12
14
16
Yellow Blue Red Gr een Yellow Blue Red Green
Lightness Pr efer ence
N
u
m
b
e
r

o
f

v
o
t
e
s

Subjective Preference.
Although this method introduced a considerable amount of data noise since the
preferences for some subjects were much more defined than for others, we believed
this was the most recommendable way to proceed as opposed to only taking into
account those participants with a clear Preference choice. In any case, if we still found
any statistically significant results that would add strength to our conclusions. Mean
reaction times and error rates for the Lightness analysis and the Preference analysis
can be seen in Table II.
Table II. Experiment 1. Mean reaction times and error rates (in parenthesis) per
experimental condition for the Lightness analysis and the Preference Analysis.
Lightness analysis Preference analysis
Position Light Dark Light Dark Preferred Disliked Preferred Disliked
Top
715.29
(6.7%)
720.71
(7.4%)

730.49
(10.6%)
716.64
(8.9%)

707.69
(6.1%)
737.14
(7.8%)

728.68
(10.8%)
726.38
(9.8%)
Down
727.40
(7.7%)
729.33
(7.7%)

738.56
(9.8%)
719.32
(9.7%)

719.90
(6.6%)
741.38
(8.6%)

732.91
(10.8%9
729.33
(10.0%)

Lightness analysis.
228
A mixed ANOVA was carried out with the factors Valence x Position x color
Lightness. Only the lightest and darkest rated colors were introduced in the analysis.
Data was again averaged across participants and repetitions for the item analysis, and
across items for the subject analysis. As shown in Graph 4 responses to negative
words were not different from responses to positive words (I: F<1; P: F<1). When the
words were displayed in the darkest color, responses were faster than when they were
presented in the lightest color, although this was significant only for the items analysis
(I: F
(1, 34)
=5.48, p=.025; P: F<1). This effect was modulated by word Valence (I: F
(1,
34)
=13.52, p=.001; P: F
(1, 34)
=6.48, p=.015), since negative words were categorized
faster when colored in dark colors (I: F
(1, 34)
=18.11, p=.001; P: F
(1, 34)
=6.74, p=.014,)
while positive words were categorized slightly faster when colored in light colors,
although the difference was clearly not significant (I: F>1; P: F
(1, 34)
=1.93, p=.174). Last,
location influenced reaction times (I: F
(1, 34)
=3.45, p=.072; P: F
(1, 34)
=4.38, p=.044) and
words presented on the top half were categorized faster. However, it did not modulate
Valence or Lightness.
650
700
750
800
Posit ive Negat ive Pos it ive Negat ive
Up Dow n
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)
Light Dark

function of position and valence. Error bars denote standard error of the mean.
Results show that perceived lightness does influence responses to word
categorization although this effect seems to only be present for negative words. It could
be the case that negative words are more susceptible to color lightness than positive
words are. However it is more probable that the effect might have to do with the colors
chosen and while the darkest colors we presented seemed to actually be coherent with
negative emotions, the lightest colors might not have been the most adequate ones.
Nevertheless, the fact that an effect was found is in itself an encouragement to extend
research and fully understand the factors that play a role in the relationship between
lightness and emotions.
229
Regarding location, the previous analysis showed that location did influence
word categorization in a different manner in negative and positive words. The fact that
this effect was not found here could be due to the smaller sample used. While in the
previous analysis all the experimental trials were included, here only those trials for
which the color present had been categorized as the darkest or lightest one were
introduced. Therefore, only half of the data is shown here. Again, such a subtle effect
might need of a large sample to show in statistical analyses.
Preference analysis.
Another mixed ANOVA was carried out taking Valence, Position and Preference
as factors. Again, only the most liked and most disliked colors were included into the
analysis. Once more, responses to positive and negative words were similarly fast (I:
F<1; P: F<1). Subjects color Preference did influence response times (I: F
(1, 34)
=12.76,
p=.001; P: F
(1, 38)
=13.81, p=.001) as words presented in the disliked color took longer to
respond than those presented in the liked color. This effect modulated Valence (I: F
(1,
34)
=20.28, p=.001; P: F
(1, 38)
=15.16, p=.001).
650
700
750
800
Up Down
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)
Liked Disliked

Graph 5. Experiment 1. Preference analysis. Mean RTs for liked and disliked colors as a
function of position and valence.
As shown by planned comparisons (see Graph 5), participants were faster to
categorize a positive word if it was colored in their preferred color (I: F
(1, 34)
=32.5960,
p=.001; P: F
(1, 38)
=30.41, p=.001) and slower to categorize a negative word when it was
colored in their preferred color although this last difference did not approach
significance (F<1; P: F<1). Location did not influence categorization times.
Results show that categorization times were also influenced by the color
preference that participants had. As found with our synesthetes, when a positive word
230
was presented in the preferred color, it was categorized significantly faster than when
presented in a disliked color. Negative words did not seem to be permeable to color
preference influence. This result is interesting in the sense that it mirrors that found in
several occasions with different synesthetes who do report experiencing discomfort
when perceiving incongruent stimuli.
Accuracy analysis.
The ANOVA with Valence, Position and Color as factors showed a main effect of
Valence but only significant for the subjects analysis (I: F
(1, 34)
=2.06, p=.160; P: F
(1,
38)
=14.53, p=.001) where negative words were more error prone than positive words.
Color also influenced accuracy (I: F
(3, 102)
=2.74, p=.047; P: F
(3, 114)
=2.55, p=.059) with
green being the color that led to more categorization mistakes. This effect was
influenced by Valence (I: F
(3, 102)
=18.41, p=.001; P: F
(3, 114)
=9.77, p=.001) due to the fact
that while negative words tended to give rise to more mistakes than positive words, this
was not the trend for green where a similar amount of errors were made independent
of word valence.
Lightness analysis.
The same ANOVA was carried out for the error rate using the recoded variable
Lightness. None of the main effects or interactions reached significance for the items
analysis. Negative words were less accurate than positive words for the subjects
analysis (I: F
(1, 34)
=1.49, p=.230; P: F
(1, 34)
=6.13, p=.018). The patterns of results
mirrored that of the RT data.
The Valence x Position x Preference ANOVA yielded a significant main effect of
Valence for the subjects analysis (I: F
(1, 34)
=2.37, p=.133; P: F
(1, 38)
=9.96, p=.003), which
was modulated by Preference (I: F
(1, 34)
=7.70, p=.009; P: F
(1, 38)
=3.76, p=.060). As can
be seen in Graph 6, planned comparisons showed that positive words were
categorized more accurately if they were presented in the preferred color (I: F
(1,
34)
=7.00, p=.013; P: F
(1, 38)
=6.12, p=.018) whereas negative words were influenced by
color Preference in the opposite direction, although not significantly (I: F
(1, 34)
=1.64,
p=.209; P: F<1). Therefore the results went in the same direction as the reaction time
data. Positive words were easier to categorize if colored on the preferred color while
negative words were not significantly influenced by color.
231
0%
5%
10%
15%
Up Down
E
r
r
o
r

R
a
t
e
s

(
%
)
Liked Disliked

Graph 6. Experiment 1. Preference analysis. Mean error percentage for liked and disliked
colors as a function of position and valence. Error bars denote standard error of the
mean.
Experiment 1 showed that the evaluation of the valence of the word can be
modified by certain factors. Specifically we found that the three variables analyzed did
show an influence on performance. Location mainly influenced positive words but not
negative words. When a positive word was presented on the upper part of the screen,
response was facilitated as compared to when it was presented on the lower half. It is
interesting that this effect was only found for positive words. Negative words were
categorized as fast when presented on the top or the bottom. This result partially
replicates Meier and Robinsons study (2004) where they found that positive words
were categorized faster when presented on the upper part and negative words
categorized faster when presented on the bottom half. Their words were presented 4
inches above fixation. At normal viewing distance 4 inches would entitle about 10
visual angle. Our stimuli were presented at 5.7; about half way. It could be that these
distances were not good enough to completely activate the up-down representation of
emotions.
Also, contrary to what would be expected, we found that red was not associated
with positive emotions since positive red colored words were the slowest ones to be
categorized. This finding clearly contradicts common sense dictating that warm colors
are associated to positive emotions while cold colors are associated to negative
emotions.
Regarding Lightness, we did find that it influenced response times. Negative
words were responded to faster if colored dark than when colored light. Positive words
were not influenced by color. If we turn to the Preference analysis positive words are
232
responded to faster if colored with the preferred color while negative words were not
influenced by coloring preference. Therefore the hypothesis about the possible
relationship between Lightness, Preference and emotions were somehow backed up
by the data. It could be concluded that positive material is susceptible to location and
color preference while negative material is susceptible to lightness influence. However,
the selection of colors used was not wide enough as to make such strong claims. It
could well be that the colors used and their specific shade interacted in such a way but
other colors would behave differently. It could also be that the effects are subtle and the
sample size was not large enough to extract such a subtle difference from the general
variability. In our view, the critical conclusion from this first experiment is that the three
variables manipulated do influence valence categorization.
Observing the voting patterns we noticed that in the Lightness ballot Yellow and
Green were the lightest colors and Red and Blue the darkest ones. In the Preference
ballot, although more distributed, Green was the most liked one and Red the most
disliked one. The same color that was most consistently liked by participants was also
a color that was most consistently though of as the lightest color. Similarly, the most
disliked color was also in the darkest color category. Except from Blue, the other three
colors seemed to follow a light-liked dark-disliked pattern.
Although Lightness seemed to only affect negative words and Preference only
positive words, when looking at the patterns of results it can be observed that the
influence of Lightness for one level of valence is usually eliminated or reversed for the
other level of valence. It is likewise for color although some of the post-hoc
comparisons are not significant). Therefore, it could be that since this is a subtle effect,
more observations were needed to eliminate noise.
One way of doing so would be to obtain data from some more participants and
another way would be to bias their responses in the color evaluations and see if that
influenced the results. In an attempt to do so we run a second experiment in which we
studied the possible relationship between color Preference and color Lightness and
how using slightly different colors would influence valence categorization. We turned
green into a dark color. If participants now chose green as the most disliked color, we
would have evidence in favor of the hypothesis that color preference is based on
lightness and independent of hue. In order to still have two darker and to lighter colors
we changed blue into a very light version of the same hue. Also, to try to maximize the
location effect we used a larger screen so that the top and bottom conditions were
even more distinctive.
233
EXPERIMENT 2. FOLLOW UP
This experiment was designed to study the influence of different lightness-hue
combinations on color Preference. We also tried to maximize the Location effect
previously found by using a larger distance between from the up and down words to
the central fixation point. Last, we expected to replicate and extend the results found in
Experiment 1.
METHOD
Participants
A different set of 42 Psychology undergraduate students participated in this
experiment in exchange for course credits. Seven of them were males and participants
mean age was 21.
The same set of 36 words was used in this experiment. Each word was
presented in one of four colors: yellow (255,255,0) and red (255,0,0) were exactly the
same as in the previous experiment. Blue was lighter (100,100,255) and green was
darker (0,100,0). The size of the screen was now 19. Therefore the words were 7.28
below or above fixation point. Everything else was kept the same as in Experiment 1.
Everything was exactly the same as in Experiment 1.
Trials were words had been incorrectly categorized (7.32%) were eliminated
from the analysis. A 2SD cut off was set to eliminate outliers. Again items (words) and
participants were used as random factor in separate analyses. A mixed ANOVA was
carried out on the a priori set of variables with 2 (Valence), 2 (Location) and 4 (Color)
as factors and the mean reaction times and error rates are shown in Table III.
234
Table III. Experiment 2. Mean reaction times and error rates (in parenthesis) per
experimental condition for the general analysis.
Valence
Positive Negative
Position Yellow Blue Red Green Yellow Blue Red Green
Top
775.52
(4.0%)
818.76
(4.2%)
797.74
(5.6%)
779.66
(3.9%)

796.27
(7.1%)
832.46
(8.3%)
797.76
(5.5%)
802.60
(8.5%)
Down
808.14
(3.9%)
849.13
(4.2%)
828.21
(6.7%)
805.73
(3.3%)

814.68
(6.2%)
861.78
(6.6%)
814.16
(5.0%)
811.16
(7.7%)

The main effect of Location was again significant, as words presented on the
upper half of the screen were responded to faster than those presented on the bottom
half (I: F
(1, 34)
=116.26, p=.001; P: F
(1, 41)
=55.99, p=.001). This location effect was again
modulated by Valence (I: F
(1, 34)
=6.90, p=.013;S: F
(1, 41)
=4.57, p=.038, see Graph 7).
When words were negative, the location in which they appeared influenced reaction
times less than when words were positive although in both cases presenting the word
on the top half yielded faster reaction times (I: both ps<.05; P: negative words: p=.078;
positive words: p=.002).
760
785
810
835
860
Negative Positive
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)
Top Down

Graph 7. Experiment 2. Reaction times as a function of valence and location. Error bars
The Color in which the word was presented also influenced categorization times
(I: F
(3, 102)
=55.18, p=.001; P: F
(3, 123)
=48.85, p=.001). Blue was now slower than any
other color (I: all ps<.001; P: all ps>.001) and all other colors did not differ (I: all ps>.1;
P: all ps>.5). This influence was modulated by Valence (I: F
(3, 102)
=3.86, p=.012; P: F
(3,
123)
=3.42, p=.019). Although blue was the slowest color, the interaction was not due to
this color but again to red. As shown by separate analysis, when only yellow, blue and
235
green were taken into account, no hint of interaction was found (I: F<1; P: F<1)
whereas when the same analysis was done now including only yellow, red and green,
the interaction was significant (I: F
(2, 68)
=5.46, p=.001; P: F
(2, 82)
=5.80, p=.005). As can
be seen in Graph 8, red colored words were slower only when positive.
700
750
800
850
900
Negative Positive Negative Positive
Up Down
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)
Yellow Blue Red Green

Graph 8. Experiment 2. Reaction times as a function of color, valence and location. Error
bars denote standard error of the mean.
The fact that blue showed the same pattern as green and yellow and also the
same pattern found in Experiment 1 for those three colors, with the only difference of a
constant 40ms increase, might be showing that this increase in reaction time for blue
was due to perceptual factors. In fact the light blue color used was slightly more difficult
to see against the grey background than the other colors and this could be accounting
for such a constant increase in RTs.
A subsequent ANOVA was again carried out in order to check the pattern of
color subjective Lightness and color Preferences. The analysis for Preference showed
that green was still highly preferred although blue was now very popular too. Yellow
obtained intermediate values and red was again the least liked color. A post hoc test
showed that only red was different from the other three colors (all ps<.05). As for the
Lightness ballot we found that now yellow and blue were categorized as the lightest
colors (no difference among them, p<.1) and red and green the darkest ones (no
difference between them, p<.5). Both light colors were statistically different than both
dark colors (all ps<.001). Therefore our manipulation was successful and participants
did perceive green as darker and blue as lighter now (see Graph 9).
236
0
2
4
6
8
10
12
14
16
Yellow Blue Red Green Yellow Blue Red Green
Lightness Preference
N
u
m
b
e
r

o
f

v
o
t
e
s

Subjective Preference. Error bars denote standard error of the mean.
An ANOVA comparing the Preference ballots for both experiments showed that
there was no interaction between Experiment and color Preference (F
(3, 213)
=1.15,
p=.331). This shows that subjects were in fact giving their Preference judgment in
Experiment 1 and it was not influenced by the Lightness of the color. Green and blue
are widely liked colors independent of their lightness. This was well complemented with
the analysis for Lightness judgment across experiments. Here we found a significant
interaction between both variables (F
(3, 237)
=56.01, p=.001) due to blue and green. Blue
was now categorized as much lighter (F
(1, 79)
=164.79, p=.001) and green as much
darker (F
(1, 79)
=52.28, p=.001).
Again a separate analysis was carried out for Lightness and for Preference and
mean reaction times and error rates can be seen in Table IV.
Table IV. Experiment 2. Mean reaction times and error rates (in parenthesis) per
experimental condition for the Lightness analysis and the Preference Analysis.
Lightness analysis Preference analysis
Position Light Dark Light Dark Preferred Disliked Preferred Disliked
Top
801.13
(4.8%)
792.71
(4.9%)

822.10
(7.2%)
790.88
(6.7%)

791.64
(3.7%)
790.24
(4.5%)

816.97
(8.5%)
791.73
(5.8%)
Down
837.47
(3.9%)
817.99
(4.8%)

841.88
(6.7%)
813.18
(6.7%)

820.36
(4.1%)
826.00
(4.2%)

837.66
(7.5%)
820.48
(6.2%)

237
Lightness analysis.
A mixed ANOVA was carried out with the factors Valence x Position x color
Lightness. Only the lightest and darkest rated colors were introduced in the analysis.
Data was again averaged across participants and repetitions in item analysis and
across items and repetitions in the subject analysis. As shown in Graph 10 responses
to negative words were not different from responses to positive words (I: F<1; P: F<1).
Responses when the words were displayed in the lightest color were slower than
responses when they were presented in the darkest color (I: F
(1, 34)
=40.25, p=.001; P:
F
(1, 41)
=15.20, p=.001). This effect was modulated by word Valence but only for the
items analysis (I: F
(1, 34)
=5.35, p=.027; P: F
(1, 41)
=1.85, p=.181). Negative words were
categorized faster when colored in dark colors (I: p=.001; P: p=.018) while positive
words were not influenced by the lightness of the color (I: p>.2; P: p>.3). Last, location
influenced reaction times (I: F
(1, 34)
=59.01, p=.001; P: F
(1, 41)
=37.21, p=.001) and words
presented on the top half were categorized faster. However, it did not modulate
Valence or Lightness.
725
775
825
875
Up Down
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)
Light Dark

All the results found in Experiment 1 were replicated here. Darker colors were
responded faster, color Lightness only influenced negative words and words presented
on the top half were categorized faster. It seems then that the effects found in the
previous experiment were not due to the specific shades used but to the relative
brightness of the colors presented since in the previous experiment the light color
category was mostly made out of green and yellow and the darkest category made out
of blue and red and in this experiment blue was now a light color and green a dark one.
238
Another mixed ANOVA was carried out taking Valence, Position and Preference
as factors. Again, only the most liked and most disliked colors were included into the
analysis. As in previous analyses words were rated as fast when they were negative or
positive (I: F<1; P: F
(1, 40)
=1.03, p=.317). Position did influence response times now (I:
F
(1, 34)
=63.67, p=.001; P: F
(1, 40)
=31.94, p=.001) and the pattern went in the same
direction as in the previous analysis (i.e. words presented on the top were categorized
faster). The main effect of Preference was significant (I: F
(1, 34)
=6.23, p=.017; P: F
(1,
40)
=6.12, p=.018) but now preferred words took longer to respond than disliked words.
This effect modulated Valence (I: F
(1, 34)
=9.32, p=.004; P: F
(1, 40)
=6.25, p=.017). As
shown by planned comparisons (see Graph 11), participants were much slower to
categorize a negative word if it was colored in their preferred color (I: F
(1, 34)
=15.39,
p=.001; P: F
(1, 40)
=11.55, p=.002) and were not influenced by color when categorizing a
positive word (I: F<1; P: F<1).
725
775
825
875
Up Down
C
a
t
e
g
o
r
i
z
a
t
i
o
n

R
T
s

(
m
s
)
Liked Disliked

Graph 11. Experiment 2. Preference analysis. Mean RTs for liked and disliked colors as a
Therefore, we found some apparent contradictions with the previous
experiment. Location showed up to influence reaction times here as it did in
Experiment 1. However, the interaction with Valence was somewhat more confusing
here. Even though it was reliable and we replicated the results from Experiment 1
because positive words presented on the top were responded to faster than positive
words presented on the bottom, the same pattern was found for negative words. We
expected negative words to be responded to faster if presented on the bottom part but
we found the opposite. They, as well, were categorized faster when presented on the
top. This advantage was of a smaller magnitude than the one found for positive words.
239
In any case, this result does not fit well with Meier and Robinson (2004). Nevertheless,
their paradigm differed from ours in an important point. While we just presented a
fixation point followed by the target word, they presented, prior to the target word, a cue
at fixation and subsequently at 1.5 and 3 inches away from it. This was followed by the
target word at distance 4 in. They were therefore promoting the allocation of attention
to the hemifield (upper vs. lower) where the word would appear. It could be that our
manipulation was not good enough to activate the up-down axis representation of
emotions and it just shows a general tendency to stimuli appearing on the top to be
processed faster. It might be that the spatial metaphor only acts when the Upper vs.
Lower location where the word appears is the location being attended. The faster RT to
words appearing at the upper hemifield might be taken as evidence that participants
were attentionally biased towards the upper hemifield, especially in this experiment in
which the eccentricity was increased as compared to the previous experiment.
Regarding color preference, while in Experiment 1 words in the preferred color
were categorized faster than those in the most disliked color, here we found the
opposite pattern. Also, while in Experiment 1 positive words were categorized faster
when colored in the preferred color and no influence was observed for negative words,
the contrary was found here. No color Preference was seen in positive words whereas
a significant effect emerged for negative words. Although contradictory at first, these
results could be explained by the abnormal effect of blue. Blue was found to be the
slowest color to be responded to and we suggest this was due to the increased
difficulty in reading blue colored words when shown in the neutral grey background.
If we go back to the Preference results, blue was the most liked color in
Experiment 2. Therefore, trials with blue colored words were the ones contributing the
most to the preferred color condition in this analysis. Blue was a constant 40ms slower
than yellow and green. Adding this general increase in RT could account for the
inverted main effect of Preference given that in Experiment 1 disliked colors were on
average 11ms slower than liked colors and here liked colors were on average 9ms
slower than disliked colors. Likewise, a general increase in the preferred color RT
(blue) would make RTs go up leaving disliked color RTs unaffected (darker colors). A
general increase in preferred color RT would make preferred color positive words
slower than before and therefore as fast as disliked color positive words. It would also
slower reaction time for preferred color negative words and since they were already
equal to those of disliked color negative words, they would become slower giving rise
to the observed pattern of results.
240
Accuracy analysis.
The ANOVA with Valence, Position and Color as factors showed no main effect
of Color (I: F<1; P: F>1) but a Color modulation by Valence (I: F
(3, 102)
=12.46, p=.001; P:
F
(3, 105)
=8.92, p=.001) due to the fact that positive words were usually more correctly
categorized than negative words except when they were colored red. Last, negative
words tended to be less accurate but only in the subjects analysis (I: F
(1, 34)
=1.54,
p=.223; P: F
(1, 35)
=11.78, p=.002). No other effect was significant.
Lightness analysis.
The same ANOVA was carried out for the error rate using the recoded Lightness
variable. Only Valence was significant and only in the subject analysis (I: F
(1, 34)
=1.38,
p=.249; P: F
(1, 41)
=10.71, p=.002). No other main effect or interaction reached
significance. The pattern of results mirrored that of the RT data.
The Valence x Position x Preference ANOVA yielded a significant Effect of
Valence (I: F
(1, 34)
=2.32, p=.137; P: F
(1, 40)
=18.76, p=.001) and Preference (I: F
(1,
34)
=2.81, p=.103; P: F
(1, 40)
=5.51, p=.024) but only in the subject analysis. Negative
words gave rise to more mistakes as well as preferred words. This last effect could be
due to blue words being amongst the preferred ones and also the most difficult color to
perceive because of the grey background. Also the interaction between Valence and
Preference was reliable (I: F
(1, 34)
=7.39, p=.010; P: F
(1, 40)
=3.99, p=.053).
0%
5%
10%
15%
Up Down
E
r
r
o
r

R
a
t
e
s

(
%
)
Liked Disliked

disliked colors as a function of position and valence. Error bars denote standard error of
the mean.
241
As can be seen in Graph 12., planned comparisons showed that negative words
were categorized more accurately if they were presented in the disliked color (I: F
(1,
34)
=9.66, p=.004; P: F
(1, 40)
=6.06, p=.018) whereas positive words were not influenced
by color Preference (I: F<1; P: F<1). Again, this is the same pattern we found for the
RT analysis.
Results from this second experiment seem to replicate those found in
Experiment 1. The same results were found regarding the influence of Position, color
Lightness and color Preference than in Experiment 1. Independent of the hue, lightness
influenced valence categorization. In a similar manner, independent of lightness, color
hue influenced valence categorization. Last, position of the word also affected
performance although the larger screen and therefore larger difference between both
positions didnt seem to have a clear influence on performance in the expected
direction.
GENERAL DISCUSSION
The main goal of these experiments was to study the extent to which colors can
affect non-synesthete individuals in a valence categorization task. Since previous
works had related color preference with emotions as well as color lightness (Collins,
1929; Cutsforth, 1925; Hemphill, 1996; Terwogt & Hoeksma, 1995) and we had
previously found random color congruency effects in other valence categorization
experiments (Callejas et al., submitted, Chapter 3) we wanted to test whether those
effects were due to some identifiable factor.
In order to test these ideas we modified the experiment previously used to
assess color effects on synesthetes and adapted it to the present situation. We
manipulated the color in which the words were presented as well as the lightness of
those colors and their position in the screen.
Results across experiments showed that positive words are consistently
mapped onto the top-down axis although negative words did not clearly show the
opposite effect. While negative words did in fact take longer to be categorized if
presented on the top half (as compared to positive words) they were still faster than if
presented on the bottom half. This result replicates previous findings in our lab where
words were presented in neutral color on the top vs. bottom of the screen and vocal
response was given. A significant interaction was found between position and emotion.
242
Last, our failure to completely replicate Meier and Robinsons results (2004) could be
due to the fact that prior to word presentation they repeatedly cued the location in
which the word was going to appear and therefore further testing will be necessary to
check this possibility.
In regards to lightness again we found a modulation of categorization times in
Experiment 1 and replicated it in Experiment 2. Negative words were more influenced
by lightness than positive words. This could be due to the small range of lightness we
used. Even though different colors had different lightness, extreme values such as
extremely dark blue or extremely light green (i.e. bluish-black or greenish-white) were
not used. It could be the case that those more extreme lightness levels would modulate
categorization times for positive words.
Likewise, a color preference effect was found where preferred colors aided in
the categorization of positive words (Experiment 1) and hindered the categorization of
negative words (Experiment 2). Again, these results show that subtle but consistent
color preference effects are found when non-synesthetes categorize emotional words.
It is important to note that these effects were found using an extremely low number of
colors and it is more than possible that none of the presented colors were in fact the
favorite color for many of the participants. Still, with a forced choice among 4 colors we
found these stable effects. If we tailored the experiment for each participant as we do
for synesthetes showing them a wide variety of colors and obtaining their most liked
and most disliked colors, we would predict that the effects we glanced here would show
in much larger degree.
The question now remains whether this effect found in non-synesthetes shares
any resemblance to the congruency effect found in synesthetes when categorizing the
valence of emotional words presented either in the congruent color or in the
incongruent color. Assuming that synesthetes results are due to the affective reaction
caused by the match-mismatch between the presented color and the internally
experienced photism as Callejas et al. (submitted, Chapter 3) showed, the question
would be whether color preference in non-synesthetes is also accompanied by an
affective reaction although of a much smaller magnitude. By definition, a preference
towards something implies that the valence of that something is larger than that of the
alternative choice. What makes these results interesting then is that we do not seem to
be conscious of the fact that our color preferences affect our performance. This seems
to be an implicit influence. Nonetheless, a post experimental questionnaire
investigating the subjective experiencing when categorizing words in different colors
243
was not collected so we cannot categorically state whether participants clearly felt that
some words were colored wrong according to their meaning or not. Further research
will show to what extent this influence of color Preference on emotion categorization is
similar to that found in synesthetes.
Last, regarding the proposed relationship between colors and emotions
(Terwogt & Hoeksma, 1995) we can conclude that, although general trends might be
there in the general population that point to some colors being more associated to
positive emotions than others, it seems to be the individual color preference what sets
the individual correspondences between colors and emotions. A possibility that is worth
investigating is that cultural color-emotion associations are part of our explicit
knowledge and individual color-emotion pairings are more implicit and depending on
the context and task we explicitly use the cultural correspondences or implicitly get
affected by the individual correspondences.

RESUMEN DE RESULTADOS

Resumen de Resultados____________________________________________________
247
En la primera seccin de esta investigacin estudiamos el efecto de la atencin
en la sinestesia. En el Captulo 1 utilizamos una tarea de Stroop sinestsico.
Presentamos un conjunto de letras y nmeros coloreados y pedimos a nuestra
participante sinestsica (MA) que nombrara el color en que estaba presentado cada
estmulo. Mediante la manipulacin del color presentado en la pantalla y el
experimentado por MA, para cada letra o nmero, obtuvimos condiciones congruentes
e incongruentes. Cuando, tanto el color presentado como el fotismo experimentado
eran iguales, encontramos que MA era ms rpida que en las condiciones en la que el
color no encajaba con el fotismo experimentado. Por lo tanto se encontr un efecto
Stroop.
Para averiguar si los colores reales interfieren con los fotismos y comparar la
magnitud de ambos efectos realizamos un segundo experimento Stroop, en el que MA
tena que nombrar el color del fotismo experimentado para cada estmulo e ignorar el
color en que se presentaba dicho estmulo. De nuevo el color presentado poda
encajar o no con el fotismo experimentado para cada letra o nmero. Aqu
encontramos que, de nuevo, MA era ms lenta al nombrar el color del fotismo cuando
no encajaba con el color en que el estmulo estaba presentado. No obstante, mientras
que nombrar el fotismo de un estmulo coloreado de manera incongruente la haca ser
unos 47ms ms lenta que nombrar el fotismo de un estmulo coloreado
congruentemente, cuando la tarea era la contraria y tena que nombrar el color de un
estmulo coloreado incongruentemente MA tardaba 285ms ms que cuando el color a
ser nombrado encajaba con el fotismo evocado por el estmulo. Esos resultados se
interpretaron como evidencia a favor de que la activacin del fotismo es bastante
fuerte e interfiere con la tarea de nombrar el color real en que el estmulo est
coloreado.
Este experimento se dise para que, reanalizando los mismos ensayos, se
pudiera tomar una medida de Priming Negativo (PN, Tipper, 1985). Analizando los
ensayos en los que el color de la dimensin previamente ignorada se converta ahora
en el color de la dimensin a la que haba que responder pudimos medir la capacidad
de MA para inhibir la informacin relevante y as realizar la tarea correctamente.
Encontramos un efecto significativo de priming negativo para la tarea de nombrar el
fotismo pero no para la tarea de nombrar el color. Esto es, cuando MA tena que
ignorar el color presentado en la pantalla, para responder diciendo el fotismo
experimentado, ella era capaz de inhibir de manera satisfactoria la activacin
proveniente del color presentado en la pantalla y, como consecuencia, en el ensayo
__________________________________________________________ RESUMEN
248
posterior en el que el nombre del fotismo evocado por el estmulo presente era el
mismo que el del color que se haba ignorado en el ensayo previo, su respuesta era
ms lenta. Sin embargo, cuando el fotismo era la dimensin distractora que tena que
inhibir, para responder correctamente al color presentado, MA no era capaz de inhibirlo
y por lo tanto, cuando esa misma informacin era necesaria en el ensayo posterior,
estaba fcilmente accesible (cuenta de ello fue la presencia de un efecto de priming
positivo, aunque no significativo desde el punto de vista estadstico).
En el Captulo 2 continuamos con el estudio del papel de la atencin en la
sinestesia. Siguiendo estudios previos (Smilek, Dixon, Cudahy y Merikle, 2001;
Palmeri y cols., 2002; Ramachandran y Hubbard, 2001a) diseamos una tarea de
bsqueda visual en la que nuestra participante sinestsica tena que encontrar un
rectngulo creado con una de tres posibles letras e informar de si se encontraba en
posicin horizontal o vertical. Las letras utilizadas se escogieron de manera que dos
de ellas compartan el mismo fotismo (letra M y letra N) y dos de ellas compartan
las mismas caractersticas de percepcin (M y W). Con este diseo podamos, por
lo tanto, disociar el grado relativo con que la forma de la letra ayudaba a la bsqueda y
compararlo con el beneficio obtenido por el fotismo evocado por cada letra. Por lo
tanto predijimos que, si los fotismos se experimentan antes de que la letra sea
reconocida, entonces encontrar un rectngulo echo de letras W y escondido entre
letras N y letras M sera ms fcil que buscar un rectngulo hecho de cualquiera de
las otras dos letras puesto que la W evocaba un color rosa-violceo mientras que las
otras dos letras evocaban el color amarillo. Por otro lado, si el fotismo se experimenta
tras el reconocimiento explcito de la letra, entonces no sera capaz de ayudar en la
bsqueda puesto que para el momento en que el fotismo se experimentara, la letra ya
se haba identificado y por lo tanto se haba encontrado el rectngulo. En este caso
esperaramos que un rectngulo creado con letras N sera el ms fcil de encontrar
puesto que era perceptivamente distinto de los distractores. En el primer experimento
encontramos que MA realiz la tarea igual que los participantes control. Era ms
rpida en encontrar un rectngulo cuando estaba formado por letras N y no mostr
un patrn distinto en la bsqueda de la M y la W. No obstante, tendi a cometer
ms errores cuando el rectngulo estaba formado por Ms que cuando estaba
formado por Ws.
En un segundo experimento incluimos una clave previa que informaba a los
participantes de la letra que formara el rectngulo para que supieran qu letra (o color
en el caso de MA) buscar. De nuevo encontramos que su ejecucin era mejor para la
249
N que para cualquier otra letra y era tan rpida encontrando un rectngulo formado
de Ms que de Ws. No obstante, volvimos a encontrar que era mucho ms exacta
por los conjuntos estimulares en los que tena que encontrar una N o una W que en
los que tena que encontrar una M.
Puesto que los controles mostraron el mismo patrn de resultados aunque con
una magnitud menor y, por lo tanto, las diferencias mayores entre MA y el grupo
control se encontraban con medidas de exactitud, llevamos a cabo otros dos
experimentos en los que estudiamos estas diferencias de exactitud. Mediante la
presentacin del conjunto estimular durante un intervalo temporal relativamente corto
(1000ms o 1500ms) medimos la cantidad de ensayos en los que la direccin del
rectngulo era indicada correctamente. Los participantes sin sinestesia cometieron
ms errores cuando tenan que identificar un rectngulo hecho de letras M o de letras
W y no encontraron una mejora en la condicin en la que los estmulos se
presentaban durante ms tiempo. Para MA encontramos que cuando el conjunto
estimular estaba presente durante poco tiempo, la bsqueda de un rectngulo hecho
de letras M era ms difcil que la bsqueda de cualquier otra letra. En la condicin
donde el conjunto estimular estaba presente durante ms tiempo volvimos a encontrar
que la letra M era la ms difcil de localizar. Adems, ahora la letra N era ms fcil
de encontrar que la W por lo que el mayor tiempo de presentacin del conjunto
estimular mejor de forma diferencial la bsqueda basada en la forma pero no la
bsqueda basada en el color.
Por ltimo, estbamos interesados en comparar los resultados descritos con
aquellos que se encontraran si los colores evocados por las letras estuvieran de
hecho presentes en la pantalla. En el ltimo experimento mostramos los mismos
conjuntos estimulares hasta que los participantes indicaban una respuesta, pero ahora
las letras N y M se presentaban coloreadas de amarillo y la letra W coloreada de
violeta. Al contrario de lo realizado en otros estudios (e.j. Hubbard, Arman,
Ramachandran y Boynton, 2005), tambin pedimos a MA que realizara esta versin de
la tarea. Encontramos que realiz la tarea de forma similar a los controles, aunque ella
fue en general ms rpida en responder. Esto pudo deberse al hecho de que tena
prctica extensiva con esta tarea. En particular, encontr los rectngulos formados por
la letra W ms rpido que los formados por la letra M y estos a su vez ms rpido
que los formados por la letra N. Estos resultados muestran claramente que en los
experimentos anteriores MA estaba identificando los rectngulos formados por la letra
N con ms facilidad debido a su forma. Por lo tanto, parece que por lo menos para
__________________________________________________________ RESUMEN
250
MA, el reconocimiento consciente de la letra es necesario para experimentar el fotismo
asociado a la misma y beneficiarse de ello.
En este segundo captulo tambin estbamos interesados en estudiar la
capacidad de MA para ignorar una dimensin de un estmulo que era incongruente con
el fotismo evocado por el estmulo, cuando atenda a otra dimensin que s era
congruente con dicho fotismo. Realizamos una versin mejorada del estudio de Rich y
Mattingley (2003), en el que presentamos un conjunto de estmulos jerrquicos del
estilo Navon, donde una forma global estaba hecha a base de distintas formas locales.
Manipulamos la consistencia entre la letra formando la forma global y la letra formando
la forma local y tambin manipulamos la congruencia entre el color presentado y el
fotismo evocado por cada una de las formas. Por ltimo, tambin manipulamos la
dimensin del estmulo a la que haba que prestar atencin (la dimensin global o la
dimensin local) y pedimos a nuestra participante sinestsica y al grupo control que
nombraran el color del estmulo presentado en la pantalla. Entre otros resultados
encontramos que, cuando una de las dos dimensiones estaba coloreada
incongruentemente con el fotismo evocado, independientemente de que fuera la
dimensin a la que se prestaba atencin o la dimensin ignorada, la ejecucin se vio
afectada en comparacin con los casos en los que ambas dimensiones estaban
coloreadas congruentemente. Este resultado es consistente con los encontrados en el
experimento de Stroop-Priming Negativo llevados a cabo en el captulo anterior ya que
habamos encontrado que ignorar el fotismo era ms difcil que ignorar el color
presentado. Aqu haba que ignorar el fotismo evocado por la dimensin atendida y el
evocado por la dimensin ignorada (a veces eran iguales y a veces distintos) para
poder responder al color presentado en la pantalla. Los resultados muestran que MA
no fue capaz de ello y el fotismo de la dimensin a ignorar afectaba la ejecucin
aumentando el efecto de congruencia.
Tras encontrar que la atencin es necesaria para procesar la identidad del
estmulo, para que este a su vez evoque un fotismo, pero insuficiente para ignorar la
dimensin irrelevante de un estmulo, en la segunda seccin nos interesamos por
estudiar las reacciones afectivas comentadas por nuestra particpante sinestsica as
como por algunos investigadores (Cytowic, 1993; Ramachandran y Hubbard, 2001b).
En el Captulo 3 mostramos, por primera vez, que estas reacciones afectivas
son verdaderas, automticas y difciles de ignorar y que ocurren como un efecto
secundario de la percepcin de un estmulo coloreado de forma congruente o
incongruente con el fotismo que evoca para una persona sinestsica particular. En
251
primer lugar pedimos a MA que evaluara la valencia de un conjunto de palabras
claramente emocionales o claramente neutrales como pueden ser amor, miedo o
mesa. Encontramos que aunque las palabras se evaluaban correctamente y las
positivas obtenan valores ms positivos que las negativas, el color en que las
palabras se haban presentado influa en sus juicios de evaluacin. Cuando las
palabras estaban presentadas en el mismo color en que MA las haba coloreado
previamente las evalu como ms positivas en general que cuando estaban
presentadas en un color incongruente. En un nuevo experimento se comprob si MA
haba obtenido este patrn de resultados porque no entendi las instrucciones y
tambin intentamos medir la automaticidad de estas reacciones afectivas. En esta
ocasin solo mostramos las palabras emocionales y no las neutrales y le pedimos que
las categorizara a la mayor brevedad como positivas o negativas presionando una de
dos teclas. Si la reaccin afectiva experimentada como consecuencia del encaje o
ausencia del mismo entre el color presentado y el fotismo experimentado es
automtica, entonces interactuara con la valencia de la palabra y dara lugar a ms
errores porque una respuesta rpida no dara lugar a que los procesos de control
actuaran para hacer que se emitiera una respuesta de acuerdo con las instrucciones.
Esto es exactamente lo que se encontr. Cuando las palabras positivas se coloreaban
congruentemente, MA daba una respuesta rpida y acertada. Cuando las palabras
positivas se presentaban en color incongruente era ms lenta y cometa ms errores.
Sin embargo, para las palabras negativas se obtuvo el patrn contrario. Si estaban
coloreadas congruentemente, los tiempos de reaccin eran ms largos y se cometan
mas errores. Si estaban coloreadas de manera incongruente, los tiempos de reaccin
eran ms rpidos y la respuesta ms exacta. Aunque lo nico que MA deba hacer era
informar de la valencia de las palabras y no tena que dar un juicio subjetivo, y aunque
se emitiera un sonido de retroalimentacin cuando cometa errores, incluso as, le
result relativamente difcil responder en los casos en los que la valencia de la palabra
no encajaba con la reaccin afectiva producida por la relacin entre el color
presentado y el fotismo evocado.
En el Captulo 4 estudiamos a otros tres participantes sinestsicos con el
mismo procedimiento y encontramos el mismo patrn de resultados. Mostraron un
efecto de congruencia APRA las palabras positivas y una modulacin del mismo para
las palabras negativas. Sin embargo, la tercera participante sinestsica que
estudiamos ni siquiera mostr un efecto de congruencia. Tardaba lo mismo en
categorizar una palabra emocional, cuando estaba coloreada congruentemente con su
fotismo, que cuando el color era incongruente. Experimentos posteriores mostraron
__________________________________________________________ RESUMEN
252
que presentaba un gran efecto Stroop muy similar al mostrado por MA en el Captulo 1.
Le costaba ms trabajo ignorar sus fotismos, cuando tenia que responder al color en
que se presentaba un estmulo, que ignorar dicho color para responder nombrando
sus fotismos. Sin embargo no mostr indicios de efecto de priming negativo para la
tarea de nombrar el fotismo ni para la tarea de nombrar el color. El hecho de que no se
encontrara dicho efecto indica que esta participante sinestsica no era capaz de inhibir
sus fotismos pero tampoco de inhibir el color real presentado en la pantalla. Uniendo
esto a los resultados de la tarea de categorizacin, los resultados sugieren que cuando
la tarea a realizar es difcil y se promueve el procesamiento del color (como es el caso
en las tareas Stroop), los fotismos s influyen en su ejecucin y esta influencia es
fuerte y difcil de inhibir. Sin embargo, cuando la tarea es ms difcil y el procesamiento
del color no se promueve (cual es el caso en la tarea de categorizacin emocional de
un conjunto de palabras) sus fotismos s son capaces de ser inhibidos o simplemente
no son lo suficientemente fuertes como para influir en su ejecucin.
Otro experimento se llev a cabo para comprobar esta hiptesis. Se
presentaron a esta participante sinestsica el mismo conjunto de palabras en los
mismos colores y a la estudiada en el captulo 1 y se les pidi que las categorizaran en
funcin de otra dimensin que esperamos que activara el fotismo de manera ms
potente. Les pedimos que indicaran si las palabras estaban coloreadas correcta o
incorrectamente de acuerdo a sus fotismos. En esta ocasin encontramos que en
ambas participantes se observaba un patrn de respuesta ms lento y ms errores
cuando las palabras positivas estaban coloreadas incorrectamente que cuando las
negativas lo estaban correctamente.
Por ltimo, para asegurarnos de que este resultado no se deba a una
contaminacin del hecho de que responder si es ms fcil cuando la palabra es
positiva y responder no ms fcil cuando la palabra es negativa que en las
condiciones inversas, realizamos un experimento control en el que pedimos a
personas sin sinestesia que indicaran si las mismas palabras estaban coloreadas
correctamente o incorrectamente basndose en un criterio que aseguraba que
accederan a su representacin semntica. Encontramos que los participantes control
no eran ms rpidos respondiendo si a palabras positivas y no a palabras negativas
de lo que lo eran en las situaciones opuestas. Por lo tanto, los resultados de nuestras
dos participantes sinestsicas no se deban a esta variable extraa sino a una
interaccin entre la valencia de la palabra y el encaje entre color y fotismo.
253
A continuacin estbamos interesados en explorar los factores que influan en
la percepcin de un color particular como incongruente. Estaba claro que una palabra
o letra presentada en el color en el que el sinestsico haba informado percibirla era
una palabra coloreada congruentemente. Sin embargo, en los Captulos 3 y 4
observamos que presentar palabras en negro no pareca influir a los sinestsicos
cuando era en un contexto bloqueado. Hasta este momento habamos utilizado el
color opuesto, en la rueda de color, al color congruente como color incongruente para
asegurar que era claramente distinto del congruente. Para explorar las razones por las
que el negro no pareca interferir en la ejecucin de las participantes sinestsicas
condujimos una serie de experimentos presentados en el Captulo 5.
Pedimos a nuestra sinestsica MA que realizara una tarea de evaluacin de
valencia emocional similar a la realizada en el Captulo 3, pero ahora manipulamos los
colores que presentbamos y el contexto en que se presentaban. Palabras coloreadas
en negro se presentaron en un contexto cromtico (es decir, mezcladas de forma
aleatoria con palabras coloreadas congruente e incongruentemente) para comprobar si
el estatus del negro como color neutral cambiaba y si ahora se perciba como un color
ms. Tambin presentamos las palabras en gris que es un tono ms luminoso de
blanco en un contexto cromtico y en uno acromtico para comprobar si producira los
mismos efectos que el negro. Por ltimo, aprovechando el hecho de que la percepcin
de color est basada en claves contextuales de contraste relativo entre el objeto
percibido y los que lo rodean, presentamos el mismo tono de gris pero cambiando las
claves contextuales para que se percibiera como blanco. Con estos experimentos
encontramos que el negro produca un patrn similar al de los colores incongruentes
cuando se presentaba en un contexto de color y algo similar ocurri para el gris.
Cuando se presentaba aislado, los juicios de valencia eran ms acertados que
cuando se presentaba en un contexto de color. Sin embargo, esta diferencia segua
siendo mucho ms pequea que la encontrada para el caso del negro. Por ltimo,
cuando las mismas palabras en el mismo color eran percibidas como blancas dieron
lugar a un patrn de evaluacin ms positivo en general. Este ltimo resultado subraya
la importancia de la experiencia subjetiva en la evaluacin de lo agradable que es un
estmulo.
En el Captulo 6 avanzamos un paso ms y estudiamos la fuerza de estas
reacciones afectivas. Para ello comprobamos si eran lo suficientemente fuertes como
para servir de estmulo incondicionado en un paradigma de condicionamiento clsico.
Intentamos inducir de forma implcita actitudes hacia un conjunto de imgenes (dibujos
__________________________________________________________ RESUMEN
254
de Pokemon) colocndolas en una localizacin espacial identificada con un nmero y
colorendolas de manera que encajaran con el color evocado por el fotismo del
nmero o no. Si las reacciones afectivas eran suficientemente fuertes, entonces
deberan de servir como estmulos incondicionados.
En un primer experimento no encontramos ningn efecto de condicionamiento
y eso puede deberse al hecho de que la relacin entre el fotismo evocado por el
nmero y el color del Pokemon no era suficientemente explcita. En un segundo
experimento incrementamos el nivel de procesamiento del nmero que indicaba la
localizacin en que se presentaba la imagen y tambin mostramos el nmero en color
congruente o incongruente en vez de mostrar en color slo la imagen del Pokemon.
Despus de dos sesiones de condicionamiento encontramos que las actitudes de MA
a cerca de las imgenes estaban influidas por el emparejamiento con un estmulo
incondicionado positivo o negativo y por lo tanto los estmulos condicionados positivos
eran evaluados como ms positivos que los condicionados negativamente. Por lo tanto
condicionamos un conjunto de imgenes mediante su emparejamiento con la
ocurrencia de una reaccin afectiva positiva o negativa fruto de la congruencia o
incongruencia entre el color presentado y el fotismo evocado.
En nuestro ltimo captulo experimental (Captulo 7) nos interesamos por
estudiar la poblacin no sinestsica y exploramos el patrn de resultados que
habamos obtenido para los participantes control a lo largo de los experimentos ya
comentados. Los participantes no sinestsicos mostraban un efecto de congruencia
por el cual realizaban la tarea de forma distinta en funcin de la congruencia
sinestsica. No obstante, este patrn apareca de forma aleatoria en tiempos de
reaccin o en porcentaje de errores y en unas ocasiones iba en la misma direccin
que en el grupo control y en otras en direccin opuesta. Aunque el patrn exacto
pareca aleatorio, el hecho de que este efecto se encontraba en muchos de los
estudios realizados nos llev a explorarlo con ms detalle e intentar encontrar una
explicacin para l.
Para comprobar qu variables estaban influyendo en este resultado estudiamos
el efecto de la claridad de los colores y de la preferencia de color de los participantes.
Tambin comprobamos el efecto de la localizacin espacial en que se encontraba el
estmulo objetivo. En dos experimentos mostramos a un amplio grupo de participantes
no sinestsicos un conjunto de palabras que se haban utilizado con anterioridad en
experimentos de sinestesia y emociones. Las presentamos en cuatro colores (amarillo,
azul, rojo y verde) y, tras pedir a los sujetos que las categorizaran como positivas o
255
negativas en funcin de su significado, les pedimos que nos informaran sobre cul de
los cuatro colores les gustaba ms y cul de ellos era el ms claro.
Los datos se recodificaron en funcin de la preferencia de color y de la claridad
subjetiva y se encontraron algunos datos bastante sorprendentes. Los participantes no
sinestsicos mostraron exactamente el mismo patrn que haban mostrado
anteriormente los participantes sinestsicos, de modo que la categorizacin de la
valencia de una palabra estaba influida por atributos del color en que se presentaba la
palabra. Cuando se presentaban en el color preferidolos participantes eran ms
rpidos en categorizar una palabra positiva que una negativa. De manera opuesta, si
se presentaban en el color menos preferido, los participantes categorizaban las
palabras negativas con ms facilidad que las positivas. En relacin con la claridad
subjetiva se encontr un patrn similar de resultados. En general, estos resultados
mostraban que los efectos de congruencia aleatorios encontrados en experimentos
previos para participantes no sinestsicos podran deberse a factores como la
preferencia por un color u otro.

GENERAL DISCUSSION

General Discussion________________________________________________________
259
SUMMARY OF FINDINGS
In the first section we studied the effect of attention on synesthesia. In Chapter
1 we used a synesthetic Stroop task. We presented a set of colored letters and
numbers and asked our synesthete to name the color in which they were presented. By
manipulating the presented color with that experienced by the synesthete for each
particular letter or number we arrived at congruent and incongruent conditions. When
both presented color and experienced photism were the same we found that she was
faster than in those cases were the color was not matching her experienced photism.
Therefore a Stroop effect was found. In order to test whether real colors would interfere
with photisms and compare both interference effects, we conducted a second Stroop
experiment in which MA had to now name the photism she experienced for each stimuli
and ignore the color in which it was presented. Again the color could either match or
mismatch that of the experienced photism for each stimulus. Here we found that again
MA was slower to name the photism when it did not match the presented color.
However, while naming the photism of an incongruently colored stimulus took about
47ms longer than naming the photism of a congruently colored stimulus, naming the
color of an incongruently colored stimulus took 285ms more than when the color to be
named matched the photism for that stimulus. This result was interpreted as evidence
that the photism activation is quite strong and it interferes with naming the actual color
in which a stimulus is presented.
This experiment was designed so that the very same trials could be used to
measure Negative Priming (NP, Tipper, 1985). By analyzing those trials were the color
of the ignored dimension became now the color of the to-be- responded-to dimension
we could measure how satisfactorily our synesthete was inhibiting the irrelevant
information to correctly perform the task. Here we found that a significant NP effect for
the photism naming task but not for the color naming task. That is, when she had to
ignore the presented color in order to respond to the experienced photism she was
able to successfully inhibit that activation coming from the presented color and
therefore it was more difficult to respond in the next trial where that previously inhibited
information was now the correct response. However, when she had to inhibit the
distracting photism information in order to correctly respond to the presented color she
was not successful in doing so and so when such information was needed in the
following trial it was readily accessible (therefore a non-significant positive priming
effect).
_______________________________________________ GENERAL DISCUSSION
260
In Chapter 2 we continued to study the role of attention in synesthesia.
Following previous experiments (Smilek, Dixon, Cudahy & Merikle, 2001; Palmeri et al.,
2002; Ramachandran & Hubbard, 2001a) we designed a visual search task in which
our synesthete had to find a rectangular shape made out of one of three letters and
inform of its position (i.e. vertical or horizontal). We chose the display letters in a way
that two of them shared the same photism (letter M and letter N) and two of them
shared the same perceptual features (M and W). With this design we could
dissociate the relative degree to which the letter shape aided in the search and
compare it to the benefit obtained by the photism elicited by each letter.
Therefore we predicted that if photisms are experienced prior to letter
recognition, finding a rectangular shape made out of letter W and embedded among
letters N and M would be easier than finding a rectangle made out of any other letter
since W elicited a purplish photism while M and N elicited a yellow photism. On the
other hand, if the photism was experienced after letter recognition, then it would not aid
in the search (i.e. by the time it was experienced the letter was already identified and
the rectangular shape found) and we would expect a rectangle made out of letter N to
be found faster than when it was made out of letter M or W because it was
perceptually different. In our first experiment we found that MA did not perform
differently than control participants. She was faster to find the rectangular shape when
it was made out of letter N and did not show a differential pattern for letter M and
W. However she tended to commit more errors when the rectangle was made out of
letter M than when it was made out of letter W.
In a second experiment we included a visual cue that informed participants of
the letter that would be making up the rectangle so that they knew what letter (or color
in MAs case) to look for. Again we found that her performance was better for letter N
than any other letter and she was as fast finding a rectangle made out of Ms than if it
was made out of Ws. Nevertheless we again found that she was much more accurate
for displays were she had to look for an N or a W than for displays were the target
as a rectangle made out of Ms.
Since controls showed the same pattern although with a smaller magnitude, we
conducted two more experiments to test this accuracy differences. Presenting the
display for a brief duration (1000 or 1500ms) we measured the amount of trials were
the direction of the figure was correctly categorized. Non-synesthetes made more
errors when having to identify a rectangle made out of Ms or Ws and the time the
display was presented did not aid them in the search. For MA we found that she in the
261
short display duration looking for an M was more difficult than any other letter. For the
longer display duration, M was still the most difficult one. Adding to this now N was
easier to find than W so a longer display differentially aided shape based search but
not color based search.
Last we were interested in comparing the described results with a case were
the colors elicited by the stimuli were actually present on the screen. In this last
experiment we showed until response the same displays but now letter N and letter
M were colored in yellow and letter W colored in purple. Contrary to other studies
(i.e. Hubbard, Arman, Ramachandran & Boynton, 2005) we also asked our synesthete
to run this version of the task. Here we found that she performed somewhat like control
participants did. Although she was overall faster than them (probably due to extensive
practice with this task), she found it very easy to identify the location of a rectangle
made out of Ws and easier to identify if it was made out of Ms than if it was made
out of Ns. These results clearly show that MA was performing better for letter N in
the first experiments because she was being led by the differential shape. When that
benefit was eliminated and now the letters were actually colored, she was much worse
to find a rectangular shape if made out of Ns than if made out of any other letter. It
therefore seems clear that, at least for MA, conscious recognition of the letter is
necessary to experience the associated photism and find a benefit from it.
We were also interested in finding how good was MA ignoring the dimension of
a stimulus that was incongruent with its elicited photism when attending to another
dimension that was congruent with the elicited photism. We run an improved version of
Rich and Mattingleys (2003) experiment. We presented a set of navon like stimuli
where a global form was made out of local forms. We manipulated the consistency
between the letter forming the global form and the local form and also the congruency
between the presented color and the photism elicited by either or both forms. We also
manipulated the dimension of the stimulus to which attention was being paid (i.e. global
dimension or local dimension) and asked our synesthete and the corresponding control
group to name the color of the stimulus presented on the screen. Among other results
we found that when either dimension was incongruently colored, irrespective of
whether it was the attended or the ignored dimension, performance was hindered as
compared to cases were both dimensions were congruently colored. This result is
consistent with the Stroop-NP experiment we had previously carried out since there we
found that ignoring the photism was more difficult than ignoring the presented color.
Here she had to ignore the photism elicited by the attended dimension as well as the
262
photism elicited by the ignored dimension (sometimes these were two different
photisms and sometimes it was just one) in order to respond to the color presented.
Results show that she was not successful in doing so and the photism of the to-be-
ignored dimension still affected performance giving rise to a congruency effect.
After having found that attention was necessary as to process the identity of the
stimulus to then elicit a photism but ineffective to filter out an irrelevant dimension of
the stimulus, in Section 2 we turned to studying the affective reactions reported by our
synesthete MA as well as other researchers (Cytowic, 1993; Ramachandran &
Hubbard, 2001b).
In Chapter 3 we showed for the first time that these affective reactions are true,
automatic and difficult to ignore reactions happening as a side effect of perceiving a
stimulus colored congruently or incongruently with the photism that it elicits for a given
synesthete. Here we first asked our synesthete MA to rate the valence of a set of
clearly emotional or clearly neutral words such as love, fear or table. We found that,
although words were correctly rated and clearly positive words were rated more
positive than clearly negative words, the color in which the words were presented
influenced her rating values. When words were presented in the same color that MA
had previously reported seeing them, she rated them as more positive overall than in
those cases where they were presented in an incongruent color.
A new experiment was run in order to test whether this effect had been due to
MA not understanding the instructions and also to test the automaticity of these
affective reactions. We now only showed her the emotional words and not the neutral
ones and asked her to categorize them as positive or negative as fast as possible by
hitting one of two keys. If the affective reaction consequence of the match-mismatch
between the presented color and the experienced photism was automatic, then it would
interact with the valence of the word and would give rise to more mistakes because a
fast response would not allow for control processes to come into play and align the
response to that requested. This is exactly what we found. When positive words were
congruently colored, MA was fast and accurate in her response. When positive words
were incongruently colored, she was slower and less accurate in her response.
However, the opposite pattern was found for negative words. If they were congruently
colored, reaction times were longer and more errors were made. If they were
incongruent, reaction times were faster and fewer errors were made. Even though MA
just had to inform about the words valence and did not have to give a subjective rating
and even though auditory feedback was given when errors were made, she still found it
263
relatively hard to respond in these cases were the valence of the word did not match
her affective reaction for that word.
In Chapter 4 we tested another three synesthetes and found the same pattern
of results in two of them. They showed a congruency effect for positive words and a
modulation of such effect for negative words. However, the third synesthete we tested
did not even show a hint of a congruency effect. She was as fast to categorize an
emotional word when it was congruently colored as when it was incongruently colored.
Subsequent experiments showed she presented a large Stroop effect very similar to
that found by MA in Chapter 1. It was more difficult for her to ignore her photisms when
having to respond to the presented color than in the opposite case. Ignoring the
presented color in order to respond to the subjectively experienced photism did not
seem a hard task. However, she did not show a hint of negative priming in neither the
photism naming task nor in the color naming task. The fact that no effect was found
indexes that this synesthete (PSV) was not able to inhibit her photisms but neither did
she inhibit the real color presented on the screen. Put together with the categorization
task, these results suggested that when the task is easy and color processing is
promoted (i.e. in the Stroop tasks), photisms do influence her performance and such
influence is strong and difficult to inhibit. However, when the task is more difficult and
color processing is not promoted (i.e. valence categorization of a set of words) her
photisms are either successfully inhibited or they are not strong enough as to influence
performance.
Another experiment was run to test this hypothesis. The same set of words in
the same colors were shown to PSV and MA and now we asked them to categorize the
words according to another dimension that we expected would activate the photism to
a higher level. We asked them to respond whether the words were correctly color
according to their photisms or not. We now found that both of them showed increased
latency and error rates when a positive word was incorrectly colored or a negative word
was correctly colored.
Last, to make sure this was not due to a confound where saying yes to
positive words and saying no to negative words was faster than the other two
conditions, we run a control experiment in which non-synesthetes had to inform
whether the same words were correctly or incorrectly colored according to a given
criteria that ensured they would access their semantic representation. We found that
they were not faster when responding yes to a positive valence word or no to a
negative valence word than when they had to say no to a positive word or yes to a
264
negative word. Therefore the findings of our two synesthetes were not due to this
confounding but to an interaction between the color-photism match-mismatch and the
word valence.
Next we were interested in exploring the factors influencing the perception of a
given color as an incongruent one. It was clear to us that a word or letter presented in
the color that they synesthete had chosen for it was a congruently colored stimulus.
However, in Chapters 3 and 4 we observed that presenting the words in black ink did
not seem to upset our synesthetes much when it was a blocked presentation. Up to this
moment we had used the color-wheel opposite of the congruent color as incongruent
color thus ensuring it would be clearly different than the congruent one. In order to
explore the reasons why black did not seem to interfere in our synesthetes
performance we run a series of experiments reported in Chapter 5.
We asked our synesthete MA to perform a task like the valence rating task of
Chapter 3 but now we manipulated the colors presented and the context in which those
colors were presented. Black was presented in a chromatic context (i.e. randomly
intermingled with congruent and incongruent colors) to check whether its neutral status
would change and it would now be interpreted as an incongruent color. We also
presented grey, a lighter shade of black in a chromatic context as well as an
achromatic one to check whether it would behave as black did. Last, we took
advantage of the fact that color perception is driven by the relative contrast of the
objects surrounding the perceived object and presented the same grey words but
changing the context cues so that it would now be perceived as white.
With these experiments we found that black behaved as an incongruently
colored when presented in a color context (i.e. mixed with congruently and
incongruently colored words) and something similar was found for grey. When it was
presented in isolation ratings were more accurate than when presented in a color
context. However, this difference was by no means as apparent as that found for black
words. Last, the very same colored words, when induced to be perceived as white
were categorized as more positive overall than when they were subjectively perceived
as grey. This last piece of data highlights the value of the subjective experience in the
evaluation of a stimulus as more or less pleasant.
In Chapter 6 we went a step further and studied the strength of these affective
reactions by testing whether they would be strong enough as to serve as unconditioned
stimuli for a classical conditioning paradigm. We tried to implicitly induce attitudes
265
towards a set of images (Pokemon cartoons) by placing them in a position identified by
a number and coloring them so as to match the color of the photism elicited by such
number or a different color. If the affective reactions were strong enough, they should
be able to serve in this procedure. In a first experiment we failed to find any
conditioning and this might have been due to the fact that the relationship between the
numbers photism and the Pokemon color was too subtle. In a second experiment we
increased the processing level of the number signaling the Pokemon location and we
also showed the number colored in a congruent or incongruent instead of only showing
the Pokemon image in that color. After two conditioning sessions we found that MAs
attitudes towards these images were influenced by the conditioning pairings and CS-
were rated as more negative and CS+as more positive. We therefore conditioned a set
of images by pairing them with the occurrence of a positive affective reaction or a
negative affective reaction fruit of a photism-color congruence or incongruence.
In our last experimental chapter (Chapter 7) we turned to the non-synesthete
population and explored a pattern that had been found throughout the different
experiments. Non-synesthete participants showed a congruency effect by which they
were different in their performance as a function of synesthetic congruence. However,
this effect randomly appeared in reaction time or in error rates and sometimes went in
the same direction as that of the synesthetes and most of the times in the opposite
direction. Even though the exact pattern seemed random, the fact that it emerged in
many of the studies led us to explore it in more detail and try to find some possible
explanation for it. In order to check the variables driving this result we tested the effect
of color Lightness and color Preference in participants performance. We also checked
whether showing the target stimuli in different locations would influence performance.
In two different experiments we showed a large group of non-synesthetes a set
of words that had been previously used in our synesthesia and emotions experiments.
We presented them in four different colors and after having them categorize the words
as positive or negative we asked them to gives us their preference ratings for each
color as well as their subjective perception of the colors lightness. Data was then
recoded as a function of color preference or color lightness and not as a function of
actual color presented and we found some quite surprising findings. Non-synesthetes
showed the exact same pattern that synesthetes had previously shown where valence
categorization was influenced by the attributes of the color in which the word was
presented. Then they were presented in the preferred color, participants were faster to
categorize a positive word than a negative one. On the contrary, if presented in the
266
most disliked color they readily categorized negative words but found more difficulty to
categorize positive ones. In regards to color Lightness a similar pattern was found.
These results were showing us that the random congruency effects found for control
participants in previous experiments could be due to factors such as their color
preference.
SYNESTHESIA AND ATTENTION
One question we wanted to answer with this research was related to the
influence of synesthesia in attention. Aware of the fact that general conclusions could
not be drawn from our data, we still wanted to learn how synesthesia was influenced by
attention at least in our sample. Not only would this add to the already collected data
from other single case studies in other laboratories but it could also give us information
about the features of our synesthetes that might be helpful to later understand their
results in other tasks such as the ones investigating the affective reactions when
dealing with incongruently colored stimuli.
1 Howautomatic is synesthesia?
In Chapter 1 we found that our synesthete MA showed a fairly large Stroop
effect when she had to name the ink in which a letter or number was presented when
such color was not the same one she internally experiences for that stimulus. In
Chapter 4 we found that our synesthete PSV showed the same pattern when naming
the color of the presented stimulus. Such a Stroop effect has been previously found by
a large number of researchers (Dixon, Smilek, Cudahy & Merikle, 2000; Mattingley,
Rich, Yelland & Bradshaw, 2001; Mills, Boteler & Oliver, 1999; Odgaard, Flowers &
Bradman, 1999; Wollen & Ruggiero, 1983) and it has been taken as evidence of
genuine synesthesia. Going further we also tested the ease to name the photism when
the displayed color did not match. In this case interference would indicate that the
processing of the color was hindering the response to the photism. For both
synesthetes we found that there was an interference effect (i.e. longer RTs when the
color and photism did not match compared when they did match) in this photism
naming task. Nevertheless, the effect was surprisingly smaller than that found in the
regular color-naming Stroop task (284.5ms vs. 47ms for MA and 105ms vs 41ms for
PSV).
267
A recent paper (Dixon, Smilek & Merikle, 2004) proposes that the different
patterns of results in these two experiments can account for the subjective reports of
synesthestes who inform that they experience their colors in the minds eye or out there
in space. According to Dixon et al. (2004) projectors find more difficulty in ignoring their
photisms since they are projected out there in space. Therefore, a much larger Stroop
effect is found for them when having to ignore the photism to name the color. Adding to
this, since photisms are presented out there, they are faster to name them than to
name the real displayed color. On the other hand, associators who see the color in their
minds eye do not find so much trouble in naming the presented color and ignoring their
internal photisms and show a smaller Stroop interference for the color naming task.
Also, since their photisms are experienced in the minds eye, they are more readily
able to name the presented color than their internal photism thus showing slower
reaction times in the photism naming task than in the color naming task. Neither of our
synesthetes reported experiencing their photisms out there in space even though their
performance resembled that of projectors more than that of associators.
Recently Ward, Li, Salih and Sagiv (2006) suggested that the associator-
projector distinction might be orthogonal to the higher-lower distinction made by
Ramachandran and Hubbard (2001b). Therefore a lower synesthete could also be
associator in the sense that he or she experienced her colors in the minds eye. Ward
et al. (2006) showed data from six associators and six projectors who showed a quite
similar pattern of results in some tasks such as the crowding experiment
(Ramachandran & Hubbard, 2001a) that was thought to provide a stronger evidence
for the perceptual nature of synesthesia than other tasks such as the embedded figures
test (Ramachandran & Hubbard, 2001). However, this hypothesis needs a set of
experiments that we did not perform in order to collect evidence in favor of our
synesthetes being what Ward would call lower-associators or higher-associators. So
far, independent of which system to categorize synesthetes proves to be more useful,
we can only say that although our synesthetes report experiencing color in their minds
eye, their behavioral data suggest that they experience a quite strong and automatic
synesthesia that leads them to show a faster and more impermeable to interference
performance when dealing with their photisms than when dealing with the visually
perceived colors.
268
2 Can synesthesia be suppressed?
Scant data has been collected to study whether synesthesia can be actively
suppressed. A way to study this is by means of the Negative Priming paradigm (Tipper,
1985) where a set of two stimuli is shown and participants have to respond to one of
them while ignoring the other one so that it does not affect their performance. In the
following trial, the to-be-responded-to stimulus might either be different from the
ignored one in the previous trial (control trial) or might be the same as the previously
ignored stimulus (ignored repetition trial). It is usually found that participants are slower
to respond to the target when it had been previously ignored and this result is
explained alluding to the fact that in order to ignore the stimulus in the previous trial, it
was inhibited thus being it more difficult to activate it now to the level of response in the
present trial. Another way to do so is by asking participants to attend to one dimension
of a stimulus and ignore the other dimension and test how the information presented in
the ignored dimension affects performance. One way to do so is by means of
hierarchical images where a larger stimulus is made out of smaller stimuli (i.e. Navon
like stimuli; Navon, 1977) and participants are asked to attend one dimension while
ignoring the other. The global precedence effect usually found in these experiments
assumes that the processing of the global level of a hierarchical pattern precedes that
of the local level. One study using each paradigm has been reported in the synesthesia
literature.
Odgaard, Flowers and Bradman (1999) carried out a Negative Priming
experiment in which they presented a synesthete with a list of colored digits and asked
her to read aloud the color in which the digits had been printed. Each color was
incongruent with her photism for the digit. They found an increased time to identify and
name the color of the targets when she had to pronounce the color name that
designated the color of her photism for the previous incongruently colored digit. Based
on this NP effect, Odgaard et al. (1999) concluded that the photisms elicited by stimuli
such as numbers can be inhibited to a considerable level when they interfere with the
task.
Rich and Mattingley (2003) reported a study using the navon-type stimulus in
which they tested and averaged a set of 14 synesthetes while they performed a color
naming task. In a first experiment, participants were requested to name the color of a
hierarchical stimulus while the congruency between the presented color and the
photism elicited either by the global, the local or both levels was manipulated. In a
269
second experiment they asked participants to perform the same task but now they had
to do it while paying attention to either the global or the local form. They computed the
congruency effect as the difference between those trials where both the global and
local form elicited the same photism as the color used to display the stimulus and those
trials where the displayed color only matched the photism elicited by the stimulus in
one of the two dimensions, either the global one or the local one. They subsequently
compared the congruency effect found in the first experiment with that found for the
same conditions in the second experiment. They found that, when attention was paid to
the global form and the color presented was incongruent with the photism elicited by
the local form, the congruency effect was no different from that found in the first
experiment. However, when attention was paid to the local form and the photism
elicited by the local form was different than the displayed color, the congruency effect
diminished as compared to the first experiment. This was taken as evidence that
actively ignoring the global form aided in the suppression of the interference coming
from the photism elicited but that dimension of the stimulus
We tried to improve both paradigms and tested MA on both of them and PSV on
the negative priming one. In Chapter 1 and Chapter 4, by presenting trials pseudo-
randomly ordered in the color naming and photism naming Stroop experiments we
were able to reanalyze that data as a function of the previously presented stimuli and
measure negative priming, not only in the regular Stroop task but also in the reversed
Stroop task. This enabled us to obtain measures from the same synesthete regarding
her ability to suppress information related to the visually presented colors or the
internally perceived photisms. MA showed a negative priming effect in the photism
naming task but not in the color naming task. That is, she was able to inhibit the
distracting information coming from the presented color when she had to name the
internally perceived photism. However, we did not find any negative priming but a non
significant positive priming trend in the color naming task. Hence, she was not able to
suppress information coming from the irrelevant photisms. This result suggests that, at
least for some synesthetes, photisms are difficult to suppress even when doing so
would aid performance. However, our second synesthete PSV who also showed a
quite apparent Stroop and reverse Stroop effect showed no hint of negative priming
neither for the Stroop experiment nor for the reversed Stroop experiment. Whereas the
differences between Odgaard et al.s (1999) results and our results for MA could be
accounted for on the basis of methodological differences (see Chapter 1 for a more
detailed discussion), the same did not apply for MA and PSV. It could be suggested
270
that PSV has a defective control mechanism and thus is not able to inhibit interfering
information.
MA also run a modified version of the global-local experiment in which we tried
to improve some of the methodological downsides of Rich and Mattingleys (2003)
design. Here we found that when both dimensions of the stimulus were congruent with
the presented color (i.e. both the global and the local letter elicited the same photism
as the color presented) as expected, MA was quite fast in her response. When both
levels induced a photism different from the presented color she was considerably
slower to name the color. Also as expected, when the attended dimension elicited a
photism incongruent with the presented color MA was as slow as when both
dimensions were incongruent. Critically, when the ignored dimension was incongruent
with the presented color the task was as effortful as when both dimensions were
incongruent. That is, she was not able to ignore the irrelevant-interfering photism
elicited by the to-be-ignored dimension.
Our results show that, at lest for MA, the photisms are very difficult to ignore
even when doing so would aid in her performance. Previous results (Odgaard, et al.,
1999; Rich & Mattingley, 2003) point in a different direction. Information about
Odgaards synesthete is not available and neither is it about Richs fourteen
synesthetes. Given the variability of results found by those researchers who take the
time to analyze individual synesthetes although they may group them on the basis of
some objective factor, it is more than probable that out of the 14 synesthetes tested by
Rich and Mattingley (2003) a good amount of them showed the described patter but
some others showed a performance more similar to that of MA. Individual differences
between synesthetes have been highlighted repeatedly (Hubbard et al., 2005; Dixon et
al., 2004; Dixon & Smilek, 2005) and until we reach a comprehensive understanding of
the factors determining synesthesia and its different manifestations, we can not discard
them as idiosyncratic features since they may hold the key to the understanding the
different varieties of synesthesia.
3 Is synesthesia preattentional?
An extensive debate regarding the necessity of attention for synesthesia has
filled journal pages in the last half decade (see Chapter 2 for a more detailed review).
Here we tried to answer this question for our synesthete MA with a visual search task
where the target was an embedded rectangle that could be either horizontal or vertical.
By manipulating the perceptual characteristics of the possible target letters as well as
271
the photism they elicited we were able to compare the benefit obtained from looking for
a perceptually salient stimulus (looking for an N among Ms and Ws) with the
benefit obtained when looking for a synesthetically salient stimulus (looking for a W
eliciting purple among Ns and Ms eliciting yellow).
Across a set of four experiments we consistently found that looking for a
rectangle made out of Ms yielded more errors than any other condition while a
rectangle formed out of Ns seemed to obtain some benefit due to its shape. If
photisms are elicited prior to conscious perception of the inducer we would expect MA
to perform much better than controls for the W condition. However, if prior
identification of the elicitor letter is necessary in order to experience the color, then N
would be the easiest condition since perceptual features made it easier to detect. The
results we found would suggest that for MA photism experience only happens after the
stimulus has been consciously perceived and therefore, it does not aid in the search.
However, the fact that looking for a rectangle made out of Ms consistently yielded
more errors than looking for one made out of Ws does not easily accommodate to this
conclusion.
By using Duncan and Humphreys (1989) explanation of visual search results
on the basis of target to non-target similarity and non-target to non-target similarity we
were able to explain the results shown by MA. Following this theory to explain the data
we can conclude that, even though photisms are not experienced prior to the conscious
perception of the shape, they do play a role in the visual search task by influencing the
similarity between target and distracters and among distracters themselves. Again,
many of the inconsistencies in the published data might be due to the specific features
of each synesthete and we can only conclude that in our case, MA does not show any
evidence for pre-attentive photism experience.
SYNESTHESIA AND EMOTION
The main aim of our work was to study the claims of synesthetes about
disliking stimuli that were not perceptually consistent with their subjective experience
for them. Since no research had been previously carried out on this topic, the questions
we wanted to answer started out at the most basic level. That is, are these reported
affective reactions actually there?. From here we built up knowledge based on the
previous step taken and the results obtained in our experiments. In the following we
provide some tentative answers to the questions we tried to solve.
272
1 Are there affective reactions associated to the experience of synesthesia?
First we wanted to test whether claims made by synesthetes and collected by
researchers (Cytowic, 1993; Ramachandran & Hubbard, 2001) about affective
reactions when perceiving incongruent stimuli were actually observable and
reproducible in a laboratory setting.
In Chapter 3 (Experiment 1a) we tested our synesthete MA and asked her to
rate the valence of a set of positive, neutral and negative words in a 7 point scale. Even
though the words were presented in color, we told her that the color was completely
irrelevant for her task and she was to judge the valence according to the semantic
meaning of the word. The colors we used to present the words were chosen so that in
half the trials they matched the photism previously reported by MA for such word and in
the other half they were the color-wheel opposite of her chosen color. Therefore we
had trials were the color presented was congruent with the photism experienced and
trials were the color and the photism were incongruent. MA showed the expected
pattern where the relative difference between positive, neutral and negative words was
found. In spite of this and although we had instructed MA to ignore the color and
respond to the semantics, we found that her ratings were overall more negative when
words were presented in an incongruent color independent of the semantics of the
word and overall more positive when the words were presented in the congruent color.
In Chapter 4 we briefly reported the results of another two synesthetes when
performing the same task but under time pressure conditions. Again we found that
although giving an accurate judgment of the words valence, this rating was influenced
by the color in which the word was presented. When asking participants about their
performance, they reported that they had followed the instructions but that some words
were just wrong.
It is crucial to note that this effect is not comparable to the congruency effect
found in Stroop experiments (Lupiez & Callejas, 2006; Hubbard & Ramachandran,
2005 for a review) where the color of the stimulus is the dimension to which participants
have to attend and respond and where the stimuli used are as simple as digits or
letters. Here, the presented stimuli are complex words that have to be semantically
processed in order to give an answer and, more important, the color in which the word
is presented is completely irrelevant for the task to be performed. Therefore, an
influence of color on the answer indicates that color is processed, a photism-color
matching process is taking place and the result of such process is in turn modulating
273
the otherwise accurate rating of the emotional words. Thus, this influence of color on
the present valence judgment task is attributed to the affective reaction produced by
the mismatch between perceived color and subjectively experienced photism.
2 Are these affective reactions automatic? Can they be ignored?
We carried out a set of experiments (Experiment 2a and 2b of Chapter 3 and
Experiment 1a and 1b of Chapter 4) to test this. Participants now had to categorize
words as being positive or negative by pressing one of two keys as fast as possible. If
the affective reaction result of the external stimuli vs. internal experience comparison
and evaluation happens in an automatic fashion, when synesthetes are requested to
give a fast response while maintaining a low error rate, this affective reaction will
interfere with the task at hand and will therefore influence performance. If such
assessment is a more conscious process that takes place later in processing, then it
should not influence a fast response aimed at a different dimension of the stimulus.
We found that MAs categorization was influenced by the match-mismatch
between the presented color and the induced photism in a revealing way. When
positive words were congruently colored she was faster than when positive words were
incongruently colored. This was not a surprise. However, when negative words were
congruently colored, she was now slower than when they were incongruently colored. A
reversal of the congruency effect was found. Again, the congruency effect and the
reversed congruency effect were caused by an irrelevant dimension of the stimulus. A
very similar pattern was found in experiment 1b (Chapter 4) for two other synesthetes
(EL and PR). EL showed a reduced congruency effect for negative words (i.e.
congruently colored words were more difficult to categorize if they were negative) and
PR showed an absence of congruency effect for the negative words.
These results can be readily explained if we assume that synesthetes are
carrying out two different evaluation processes. On the one hand they are voluntarily
accessing the meaning of the target word and assessing whether it has a positive or
negative connotation. On the other hand they are automatically, and perhaps
involuntarily, assessing the fit between the externally perceived stimulus and the
internally experienced color sensation. This second assessment interacts and
influences the final response. If the result goes in the same direction as that of the
words connotation, the summation of both results aids in the execution of a fast
response in the correct direction. That is, if a positive word is shown in a congruent
color, the semantic assessment voluntarily carried out will signal for a positive
274
response. The automatic match-mismatch computation will also give a positive result
because the match between the perceived color and the internal experience is good.
Evidence from both processes will add up and a response will be executed. However, if
the positive word is incongruently colored, while the access to the words semantics will
be unchanged, the outcome of the match-mismatch computation will be of a different
sign. A negative outcome will now emerge since the perceived color does not match
the internally experienced one. When the result from both computations merges a
conflicting situation is created and it requires effortful control to be solved. Thus longer
reaction times are found.
Crucially, this hypothesis predicts that a negative word incongruently colored will
be responded to faster than when congruently colored. The semantic assessment will
analyze the words meaning and give a negative response as a result of such
process. In a parallel fashion, the match-mismatch computation will inform that both
experiences are contradictory and will therefore send a negative response as a result.
Again, information coming from these different processes will accumulate and a fast
response will be executed. Again, when the negative word is congruently colored, the
information coming from both processes will be of different sign and control will have to
be exercised in order to solve it and give an accurate response.
Our data from the error rates also show the same pattern described here. MA
was far more error prone for those inconsistent conditions (i.e. positive incongruent
words and negative congruent words) than for those conditions where both dimensions
went in the same direction (i.e. positive congruent words and negative incongruent
words). EL also showed a large congruency effect for positive words and an inverted
pattern for negative words. PR was very accurate and those few mistakes she made
showed no differential error rate pattern for any of the conditions.
The fact that a large percentage of errors was found, especially large for MA
helps us in addressing our second question. The task to be performed was fairly simple
because we used words that had already been rated by a large sample and were
clearly positive and negative and even the synesthetes had already been exposed to
them in the rating experiment and had a clear opinion about their connotation. Adding
to this, feedback was given for inaccurate responses. Considering all these facts, it is
surprising to find such a high percentage of mistakes especially for MA but also for EL.
We suggest that a considerable control effort was being made to correctly respond but
the affective reaction resulting from the match-mismatch could not be successfully
275
suppressed and it still influenced behavior. The absence of an effect for PR in error
rates could point to a better control mechanism.
Again, large differences could be found between different synesthetes and while
some might be able to ignore this affective reaction elicited by the match-mismatch
between external and internal experiences, some others can not control them and this
shows in their performance.
3 Are affective reactions a ubiquitous phenomenon?
If possible, we were also interested in answering the question of how common it
is to experience affective reactions associated to a color-photism mismatch. Out of the
four synesthetes tested, one of them PSV showed no hint of a congruency effect in the
word categorization and valence rating experiments (Chapter 4, Experiment 1a)
although she did show a large Stroop and reverse Stroop effect (Chapter 4, Experiment
2a and 2b). Further testing was carried out to check whether the lack of an effect was
due to an absence of synesthesia for words or to a weaker photism for them. The fact
that PSV reported realizing that some words were wrongly colored suggests that she
did experience her photisms for the stimuli. In Experiment 3 (Chapter 4) we asked her
and MA to categorize the very same words as correctly or incorrectly colored according
to the photism they experienced for them.
By promoting the processing of the photism we found that MA and, most
important, PSV both showed an interaction between color-photism congruency and the
valence of the word that was now the irrelevant dimension of the stimulus. That is, the
word was processed to a high extent (at least enough as to induce the activation of the
corresponding photism although probably more) and the color in which the word was
presented had to be compared to the internally experienced photism. Now when these
three dimensions were processed we found that, again, those cases where both
assessments yielded a result in the same direction (i.e. positive correctly colored words
or negative incorrectly colored words) were responded to faster than the cases where
mixing results were coming from one and another assessment even when this meant
congruently colored words taking longer to be categorized as colored right.
These results suggest that Mattingley and Rich (2004) might be right when they
suggest that the difference in the strength of the synesthetes photisms could be the
factor underlying the variability among synesthetes in different tasks. Although we do
not agree with them when stating that this is the only factor that explains the different
276
patterns found for different groups of synesthetes, we do think that it might be one of
the influencing factors that explains the lack of a congruency effect in PSV when
having to give an emotional categorization of a set of words. We also suggest that
another variable that could be playing a role has to do with personality traits such as
anxiety. It has been shown that one of the defining features of anxious people is a lack
of an efficient control mechanism to deal with the ongoing events (Lazarus, 1991).
Anxious people would be less able to deal with conflicting situations and control those
dimensions of the stimuli they are dealing with that interfere in their performance. We
found a nice correspondence between anxiety trait measures (STAI) and our
synesthetes performance on the emotional categorization task (Experiment 2a in
Chapter 3 and Experiment 1a in Chapter 4). MA showed a high influence of the color-
photism match-mismatch on her valence categorization times and she also showed
quite high levels of anxiety (percentile 80). On the other hand, PSV did not show a hint
of a congruency effect in this task and this correlated with very low levels of anxiety as
a trait (percentile 11). It could well be that MA is not able to control the irrelevant
information coming from the color dimension when performing this task while PSV
concentrates on the task required and is able to completely control the influence of the
perceived incongruence on her performance.
This discussion merges with that of the previous section and points some of the
conditions that might play a role in the ability to suppress or at least ignore the affective
reaction coming from the photism-color incongruence. As more synesthetes are tested
in these tasks we will be able to explore this and other factors that could be explaining
such individual differences found here.
4 Can these affective reactions be modified or extinguished?
A fortuitous observation made us realize that while rating a set of black words
was not different between our synesthete MA and her control group if they were
presented in a blocked fashion, when the very same words were presented randomly
mixed with congruently and incongruently colored words they were rated in a different
way by MA. Her pattern for black colored words resembled now the one found for
incongruent words (Chapter 3, Experiment 1a and Chapter 5, Experiment 1).
We asked ourselves about the mechanism subtending this pattern of results. If
MA was now rating the very same words in the very same color differently, either she
voluntarily decided to do so or something about the situation involuntarily influenced
her. The fact that across experiments she was always showing a valence effect (i.e.
277
more positive ratings for positive words than for negative words) made us think that she
was actually performing the task according to the specified instructions and not based
on her subjective approval or disapproval of the presented stimuli. Having discarded
the first possibility, we now faced different alternative involuntary mechanisms that
could be influencing MAs ratings.
The starting point for a synestheticaly affective reaction is a) a stimulus eliciting
a synesthetic perception, b) the stimulus being inconsistent with the synesthetic
perception and c) a comparison process that checks the match-mismatch between the
perceptual and the synesthetic dimension and outcomes a right vs. wrong verdict.
Since the stimuli were the very same ones in both mixed and blocked situations, the
stimulus was inconsistent with the synesthetic perception, that is, the second factor
was also present. Thus, the results found could be due to one of two different
possibilities. Either the stimuli were not eliciting a synesthetic perception when shown
in a blocked presentation, or the comparison process was not performing properly
under these circumstances. Last, a third possibility was that the comparison
mechanism was active but its outcome was somehow inhibited so as to not affect
performance. Again the first option did not seem too plausible considering the data
patterns found throughout all our investigations. MA showed a considerably strong
synesthesia and its influence on performance is difficult to inhibit as seen in the
Negative Priming experiments, the hierarchical figures task or the valence
categorization studies (Experiment 2 and 3 in Chapter 1, Experiment 5 in Chapter 2,
Experiment 2a in Chapter 3, etc.).
Experiments 2a, 2b in Chapter 5 showed that the second or the third options
were more plausible. That is, the comparison process either was not performing
properly or was being inhibited under some circumstances. In Experiment 2b we found
that presenting the same words all colored in grey resulted in a rating pattern very
similar to that found for black words when mixed with colored words. In the same
manner, when grey was intermingled with congruently and incongruently colored
words, the pattern found treated grey words as very incongruent. It seems very
improbable that the very same word would elicit a photism if presented in black in a
color context but not when presented in an achromatic context. It becomes even more
improbable taking into account that grey, a lighter version of black, did not behave in
the same manner and behaved in a way very similar to incongruently colored words
irrespective of its context.
278
Last, extensive questioning as well as Experiment 3 in Chapter 5 hinted in the
direction of the comparison process taking place and the outcome being somehow
more or less ready to influence behavior. The very same words in the very same grey
color, when presented in a context of darker colors were interpreted by MA as being
white. When they were perceived as white and were presented in an achromatic
context, their evaluation still resembled that of incongruently colored words. However, it
was significantly less negative than that given to the grey words presented in an
achromatic context. Analyses of MAs reports suggest that black ink in an achromatic
context has a special status and it is due to intensive and extensive practice with
reading print in black ink. If books and computers printed by default in green, for
example, it is very probable that reading a uniform green page was not found aversive
by a synesthete the same way that no reports of reading a black printed page being
aversive have been found. This explains why grey behaved as an incongruent color
even when shown in isolation.
Nevertheless, our results were not able to clearly differentiate between the last
two possibilities. In order to discern whether the comparison does not take place or is
somehow inhibited, other paradigms would be more conclusive. Event related
potentials measured while performing a speeded categorization task might be a good
way to study this. If the outcome of the comparison process is somehow suppressed
but not completely inhibited we might find a more subtle version of the pattern we
would normally find for incongruently colored words. If the evaluation process does not
even take place, the pattern of brain activity would presumably show topographic or
component differences.
Last, the proposal that the comparison process is still taking place and its
outcome does not exert a complete influence on the subsequent processing and
further, the suggestion that this is due to extensive practice, inevitably leads to the
prediction that any given color can become a neutral color through practice and give
rise to a pattern similar to that found for black ink when presented in isolation.
Obviously ethical matters make it difficult for such a study to be conducted.
5 Are synesthetic affective reactions strong enough as to condition co-occurring
events?
Previous research has shown that positive and negative valence words can
serve as unconditioned stimuli in a classical conditioning procedure and they can
influence the posterior attitude towards otherwise neutral stimuli they are paired with
279
(Olson & Fazio, 2001). In this study they used normalized images and pre-tested words
as US. Therefore both types of US were clearly positive or negative. They obtained,
after conditioning, a 0.64 differential evaluation score between the CS+and the CS-
with a sample of 45 participants. We attempted to use the postulated affective reaction
result of the incongruence between presented color and experienced photism as the
US stimulus in our procedure. Therefore when a number signaling a spatial position
was presented in a color incongruent with its photism, a negative affective reaction was
expected to occur. Conversely, if the number indexing the location was colored
according to the elicited photism, a positive affective reaction was expected. By
consistently presenting a neutral stimulus in one of these two conditions we expected
to condition it with the affective reaction taking place so that after conditioning it would
turn to be perceived as positive or negative instead of neutral. We found a difference
between CS+and CS- of 1.8 (-0.31 vs. 1.49).
These results clearly suggest that the affective reaction experienced when a
color-photism mismatch is perceived is of a strong nature and can account for effects
such as those described here. A neutral stimulus can acquire a positive or negative
connotation as a function of the stimuli that co occur with it. Although no test was
carried out to check the decline of this effect over time, it was presumably quite fast.
Nevertheless, the crucial fact is that, if a simple task like the one carried out here is
able to condition an attitude towards a previously neutral stimuli in less than thirty
minutes, a life-long series of synesthesia experiences interacting with objects in the
real world that may or may not show a correspondence with the internally perceived
colors could be a potentially strong determinant of synesthetes attitudes towards
different objects, persons or situations.
6 Proposal of the bidimensional assessment hypothesis
Taking into account the results from the several series of experiments, it can be
concluded that synesthetic affective reactions seem to be strong enough as to have an
influence on valence categorization and rating of emotional words, and even to
condition neutral stimuli, thus influencing attitudes towards them. In the following we
suggest a model that could account for these results. We propose that, a comparison
mechanism that we term synesthetic assessment (SynA) is continuously checking for
inconsistencies between both perceptual and synesthetic experience. Some factors
modulate whether the results of this synesthetic assessment influence performance or
not.
280
We suggest that when an experimental task has to be carried out as in our
studies, the SynA modulates the result of the task being performed by adding evidence
that biases the outcome of the required task. Specifically, when participants had to
perform a semantic categorization task regarding the connotative meaning of a set of
words, a semantic assessment process (SemA) checks the meaning of the word and
sends information about the words connotation. In our tasks where positive and
negative valence words were used, this SemA process informs whether the semantic
dimension of word is positive or negative. At the same time, the SynA process informs
about the synesthetic dimension of the stimulus and its relation to the external world.
Its outcome can be positive (i.e. when internal and external features match) or negative
(i.e. when internal and external features do not match). In a later stage, information
from both assessment processes is compared and ultimately, the result of this
comparison is sent to the response mechanisms.
As shown in Figure 1, when a positive word is presented in a congruent color,
the outcome of the SemA is of a positive sign and the result of the SynA is also of a
positive sign. When the same positive word is now presented in a different color, while
the SemA is still producing the same outcome, the SynA is now informing of a negative
result: the perceived color does not match the internally experienced one, and thus a
negative affective reaction is elicited. When in the next stage of processing these two
sources of information are combined, the result can be a coherent one (both processes
yielding the same outcome) or an incoherent one (each process yielding an outcome of
different sign). For consistent outcomes information is sent to the execution modules
and in turn a response is performed. For inconsistent outcomes a control mechanism
has to bias the competition so that the outcome of the stimulus dimension to which the
participant has to respond to is the one that ultimately gets relayed to the execution
modules. Therefore, more errors and longer reaction times are expected under these
circumstances.
This model predicts that for negative words presented in a congruent color, the
response will resemble that of incongruently colored positive words. Although intuitively
it could be thought that since words are colored congruently they are processed faster,
this model proposes that, given the automaticity of word reading, what is crucial is the
coherence between information coming from different assessment processes. For
example, in a negative congruently colored word, the SemA yields a negative outcome
while the SynA yields a positive outcome. A conflicting situation is encountered and
control processes have to bias the competition. Therefore, this model would propose
281
that negative incongruently colored words would be easier to respond to because both
SemA and SynA provide similar outcomes.
BI-DIMENSIONAL ASSESSMENT HYPOTHESIS
Context

Figure 1. Depiction of the Bidimensional Assessment Hypothesis proposed to explain the
affective reactions found in synesthetes when perceiving a stimulus colored congruently
or incongruently with their internal experience.
As shown by our experiments, the results are not as clear cut as predicted by a
simple version of the model, as different factors have been shown to influence the
outcome. In order to account for the variable results found with other synesthetes we
propose that, although the SynA process is present in all synesthetes, it might be more
active in some synesthetes and therefore influence their performance to greater levels.
Its effect can also depend on the strength and specificity of the synesthetic
associations of each synesthete.
Regarding the noisier results found for negative words we propose that again,
the strength of both assessment processes can account for this patterns. The
processing of negative stimuli seems to be fairly automatic and it could be that the
outcome of the SemA when categorizing negative words was stronger than the
outcome of the SynA. Only if synesthesia is strong enough would the SynA outcome
have the ability to influence the SemA.
Stimulus
Color
Semantic
assessment
Synesthetic
assessment
Strength of
synesthesia
SynA
output SemA
(Variable strength) output
CONTROL
MECHANISMS
Integration/competition of assessment
outputs and co-occurring events
Response
system
Modulating factors
282
Our context experiments informed that not all colors are equally incongruent
and that other factors, such as the context in which the task is performed, can also
influence performance. We propose that the SynA process does not only take into
account the photism and presented color but SynA is a more complex process shaped
by the needs of synesthetes to adapt to their environment. Thus, it is influenced by
factors such as the specific color presented or the context in which the synesthete is
(i.e. performing a task with plenty of colors, reading a printed book, etc.). In turn, this
context and color modulations have been shaped by adaptation and we propose that
their mechanism of action is to weaken the strength of the SynA outcome.
Last, the fact that attitudes can be formed as a result of pairing a stimulus with a
synesthetically incongruent situation supports the claim that this SynA process is
ubiquitous. Nevertheless, the demands of the task being performed could influence the
outcome of the process and make it more available to conscious report while in other
situations it would just be implicitly acting on the behavior without the synesthetes
noticing it.
FUTURE DIRECTIONS
These series of experiments have provided a starting point for the empirical
study of affective reactions experienced as a side effect of the synesthetic perception.
Future research based on these findings should aim at answering, among others, the
following questions.
Individual differences
As already mentioned, most synesthesia researchers agree that synesthetes
are a very heterogeneous group in regards to the inducers and concurrents they
experience and even within the same subtype of synesthesia, in regards to the way
they experience those concurrents and how that affects performance.
It is a natural deduction that the affective reactions experienced as a side effect
of synesthesia might also vary in degree among different synesthetes. We have shown
that this phenomenon seems to be consistently found in different synesthetes of quite
varied backgrounds. However, we have also shown that some synesthetes, although
reporting the experience of such reactions, do not behaviorally show it, or only show it
under certain circumstances where the affective reaction is measured quite indirectly.
283
Future research should extend these findings to a larger population of
grapheme-color synesthetes and face the challenge of studying it in other subtypes of
synesthesia such as lexical-gustatory synesthesia or music-color synesthesia. It would
also be interesting to check whether the proposed subtypes of grapheme-color
synesthesia (Ramachandran & Hubbard, 2001b and Smilek et al., 2001) show different
affective patterns when facing incongruently colored stimuli. Furthermore, it will be
interesting to investigate whether the differences among synesthetes regarding the
affective reactions elicited by incongruent stimuli are determined by the type of
synesthesia, by personality traits or by an interaction between these two factors.
Factors influencing synesthetic affective reactions
Another area of research inspected here is that of the factors that influence the
synesthetically induced affective reactions. Future research should test the influence of
the proposed factors with other paradigms and test the extent to which they give rise to
a fully differentiated behavioral pattern. Personality traits such as anxiety or self
regulation abilities as well as cognitive factors such as the efficiency of control
mechanisms should be studied together with other factors such as the amount of
exposure to different incongruent pairings.
Some synesthetes have argued that another factor influencing the perception of
a stimulus as incongruent with the internal experience is that of its authorship.
According to this reports, even a congruently colored stimulus would induce a negative
reaction because it is an artificial display presented to the synesthete. Likewise, if it is
the synesthete who voluntarily writes a correspondence letter with a green ball pen, the
inconsistencies will not be experienced as negative because it was ones own decision.
Last, it would also be interesting to check whether spatially closer hues interfere
to a lesser degree than further located hues. In our studies we used the color-wheel
opposite of the congruent color to assure an effect was found. It would be interesting to
study whether the magnitude of the effect is kept constant despite of the color
presented as incongruent or whether as that color approaches the photism color in the
color wheel, the affective reaction decreases.
Physiological measures of synesthetically elicited affective reactions
Although we showed here that there is an affective reaction associated to the
perception of incongruently colored stimuli, we did not show whether that reaction is
284
also physiological. This could be studied with the use of psycho-physiological
techniques such as skin conductance response. Pilot experiments carried out in our
laboratory suggest that this technique might be a good candidate to show convergence
of results with different approaches.
Event related potentials can also shed some light to the processes underlying
affective reactions in synesthesia and whether the absence-presence of such reactions
in the behavioral measures is correlated to quantitative or qualitative differences in
cerebral activation patterns.
Similarities between synesthetes and non-synesthetes
Last, it is also worth noting that one of the reasons why synesthesia is such an
interesting phenomenon is because knowledge about its underlying processes can
illuminate the investigation of cross-modal integration in non-synesthetes. Here we
showed that control participants produce reaction time patterns very similar to those of
synesthetes when presented with words where the relevant dimension (words
connotation) was not matching an irrelevant and probably implicit dimension (the
participants color preference). Although the sources of mismatch are different from
those present in synesthetes, it could well be that the underlying mechanisms
responsible for both patterns were shared by synesthetes and non synesthetes.
Studies bridging the gap between synesthetic affective reactions and preferences in
non-synesthetes regarding their modulation of emotional processing will be a good
attempt to increase our knowledge emotional evaluation of stimuli.
CONCLUSSIONS
Our main goal was to study the automaticity of synesthesia and whether it is
subject to inhibition and to empirically test for the existence of affective reactions
resulting from a mismatch between what a synesthete internally experiences when
confronted with a stimulus and what is actually present and sensory processed.
We found that synesthesia is a fairly strong phenomenon that happens
automatically and is difficult to ignore. Therefore, it affects performance even when it is
detrimental for the task at hand. Adding to this, we also found that an affective reaction
is experienced as part of the synesthesia phenomenon and it clearly influences
behavior. This pleasantness or discomfort feeling is experienced automatically and it is
difficult to ignore. We also found that it can be modulated by contextual features or
285
learning and adaptation. Depending on some personality traits such as anxiety and on
the control mechanisms of the synesthetes its effect on performance might be
controlled under certain circumstances when trying to accurately perform a task. Last, it
can influence co-occurring events and bias the subjective attitude towards them.

REFERENCES

References ______________________________________________________________ 289
Acosta, A., Lupiez, J ., Fresse, B. & Megas, J .L. (under revision). Emotional priming
and interference effects in a modified Stroop colour naming task in anxiety-trait
versus depressive participants.
Adams, E. & Osgood, C.E. (1973). A cross-cultural study of the affective meaning of
color. Journal of Cross-cultural Psychology, 4, 135-156.
Aleman, A., Rutten, G.J ., Sitskoorn, M.M., Dautzenberg, G. & Ramsey, N.F. (2001).
Activation of striate cortex in the absence of visual stimulation: an fMRI study of
synesthesia. Neuroreport, 12(13), 2827-2830.
Allison, T., McCarthy, G., Nobre, A., Puce, A. & Belger, A. (1994). Human extrastriate
visual cortex and the perception of faces, words, numbers, and colors. Cerebral
Cortex, 5, 544-554.
Amaral, D.G.l, Prince, J .L., Pitanen, A. & Carmichael, S.T. (1992). Anatomical
organization of the primate amygdaloid complex. In J .P.Aggelton (Ed.) The
amygdala: neurobiological aspects of emotion, memory and mental dysfunction.
New York: Wiley.
Argelander, A. (1927). Das faberhren und der synsthetische factor der
wahrnehmung. J ena, Germany: Fischer.
Asher, J .E., Aitken, M.R.F., Farooqi, N., Kurmani, S. & Baron-Cohen, S. (2006).
Diagnosing and phenotyping visual synaesthesia: a preliminary evaluation of the
revised test of genuineness (TOG-R). Cortex, 42(2), 137-146.
Asher, J .E., Lamb, J .A., Baron-Cohen, S. & Monaco, A. (2006). Ongoing results of a
whole-genome screen for susceptibility genes linked to synaesthesia. Oral
presentation at the 2
nd
international meeting of the UK Synaesthesia Association
celebrated in London, UK. April 21-23.
Azoulai, S., Hubbard, E. & Ramachandran, V.S. (2005). Does Synesthesia Contribute
To Mathematical Savant Skills? Journal of Cognitive Neuroscience, 69-69 Suppl.
S 2005.
Baron-Cohen, S. & Harrison, J .E. (1997). Synaesthesia: Classic and Contemporary
readings. Oxford, England: Blackwell.
290 _____________________________________________________ REFERENCES
Baron-Cohen, S., Burt, L., Smith Laittan, F., Harrison, J . & Bolton, P. (1996).
Synaesthesia: Prevalence and familiarity. Perception, 25(9), 1073-1079.
Baron-Cohen, S., Wyke, M. A. & Binnie, C. (1987). Hearing Words and Seeing Colors:
an Experimental Investigation of a Case of Synesthesia. Perception, 16(6), 761-
767.
Beeli, G., Esslen, M. & J ancke, L. (2005). Synaesthesia: when coloured sounds taste
sweet. Nature, 434, 38.
Blake, R., Palmeri, T.J ., Marois, R. & Kim, C-Y. (2005). On the perceptual reality of
synesthetic color. In L.C.Robertson & N.Savig (Eds.), Synesthesia: Perspectives
from Cognitive Neuroscience. New York: Oxford University Press, pp. 47-73.
Brugger, P. Knoch, D., Mohr, C. & Gianotti, L.R.R. (2004). Is digit-color synaesthesia
strictly unidirectional ? Preliminary evidence for an implicitly colored number
space in three synaesthetes. Acta Neuropsychologica, 2(3), 252-258.
Brugger, P., Della Betta, M.A. (2000). "Scarc-Effect" In Synaesthesia? Evidence For An
Analog Representation Of Colors Associated With Digits. European Journal of
Neuroscience, 12, 167-167 Suppl. S 2000.
Calkins, M.W. (1893). A statistical study of pseudo-chromatesthesia and of mental
forms. American Journal of Psychology, 5, 439-464.
Calkins, M.W. (1895). Synesthesia. American Journal of Psychology, 7, 90-107.
Callejas, A., Acosta, A. & Lupiez, J . (submitted). Green love is ugly: Emotions elicited
by synesthetic grapheme-color perceptions.
Callejas, A., Smilek, D., Dixon, M. & Merikle, P. (2004). A web-based descriptive study
of synaesthesia. Oral presentation at the 4
th
international meeting of the American
Synesthesia Association celebrated in Berkeley, US. November 2004.
Carter, C.S., Macdonald, A.M., Botvinick, M., Ross, L.L., Stenger, V.A., Noll, D. &
Cohen, J .D. (2000). Parsing executive processes: strategic vs. Evaluative
functions of the anterior cingulate cortex. Proceedings of the National Academy of
Sciences of the United States of America, 97, 1944-1948.
References ______________________________________________________________ 291
Castel, L.B. (1735). Nouvelles experiences doptique and dacoustique. Mmoires pour
lHistoire des sciences et des beaux arts.
Catena, A., Fuentes, L. & Tudela, P. (2002). Priming and interference effects can be
dissociated in the Stroop task: new evidence in favor of the automaticity of word
recognition. Psychonomic Bulleting and Review, 9(1),113-118.
Cohen, J .D., Dunbar, K. & Mcclelland, J .L. (1990). On the control of automatic
processes- a parallel distributed-processing account of the Stroop effect.
Psychological Review, 97, 332-361.
Cohen, L. & Dehaene, S. (2004). Specialization within the ventral stream: the case for
the visual word form area. Neuroimage, 22, 466476.
Cohen-Kadosh, R. & Henik, A. (2006). Color congruity effect: where do colors and
numbers interact in synesthesia?. Cortex, 42(2), 259-263.
Cohen-Kadosh, R., Sagiv, M., Linden, D.E.J ., Robertson, L.C. Elinger, G. & Henik, A.
(2005). When blue is larger than red: colors influence numerical cognition in
synesthesia. Journal of Cognitive Neurosciences, 17(11), 1766-1773.
Collins, M. (1929). A case of synaesthesia. Journal of General Psychology, 2, 1227.
Cutsforth, T. D. (1925). The role of emotion in a synaesthetic subject. American Journal
of Psychology, 36, 527543.
Cynader, M.S., Andersen, R.A., Bruce, C.J ., Humphrey, D.R., Mountcastle, V.B., Niki,
H., Palm, G., Rizzolatti, G., Strick, P., Suga, N., von Seelen, W. & Zeki, S. (1988).
General principles of cortical operation. In P.Rakic & W.Singer (Eds.),
Neurobiology of Neocortex. New York: J ohn Wiley, pp. 353-371.
Cytowic, R.E. & Wood, F. B. (1982). Synesthesia. II. Psychophysical relations in the
synesthesia of geometrically shaped taste and colored hearing. Brain and
Cognition, 1(1), 36-49.
Cytowic, R.E. (1989). Synesthesia: a union of the senses. Springer, New York.
Cytowic, R.E. (1993). The man who tasted shapes. New York: Putnam.
292 _____________________________________________________ REFERENCES
Cytowic, R.E. (1995). Synesthesia: Phenomenology and neuropsychology. A review of
current knowledge. Psyche, 2(10).
Cytowic, R.E. (2002). Synaesthesia: a Union of the Senses. 2nd Edition. Cambridge:
MIT Press.
DAndrade, R. & Egan, M. (1974). The colors of emotion. American Ethnologist, 1(1),
49-63.
Day, S. (2003). Types Of Synaesthesia. Web Site on Synaesthesia Maintained by Sean
Day. Http://Home.Comcast.Net/~Sean.Day/Types.htm.
Day, S.A. (2005). Some demographic and socio.cultural aspects of synesthesia. In
L.C.Robertson & N.Savig (Eds.), Synesthesia: Perspectives from Cognitive
Neuroscience. New York: Oxford University Press, pp. 11-33.
Dehaene, S. (1997). The number sense: How the mind creates mathematics. New
York: OUP.
Delay, J ., Gerard, H.P. & Racamier, P.C. (1951). Les synesthsies dans lintoxication
mescalinique.LEncphale, 40, 1-10.
Di Lollo, V., Enns, J . T. & Rensink, R. A. (2000). Competition for consciousness among
visual events: The psychophysics of reentrant visual processes. Journal of
Experimental Psychology: General, 129, 481507.
Dixon, M.J . & Smilek, D. (2005). The importance of individual differences in grapheme-
color synesthesia. Neuron, 45(6), 821-823.
Dixon, M.J . Smilek, D., Duffy, P.L., Zanna, M.P. & Merikle, P.M. (2006). The role of
meaning in grapheme-colour synaesthesia. Cortex, 42(2), 243-252.
Dixon, M.J ., Smilek, D. & Merikle, P.M. (2004). Not all synaesthetes are created equal:
Projector versus assoicator synaesthetes. Cognitive, Affective and Behavioral
Neuroscience, 4(3), 335-343.
Dixon, M.J ., Smilek, D., Cudahy, C. & Merikle, P.M. (2000). Five plus two equals yellow.
Mental arithmetic in people with synaesthesia is not coloured by visual
experience. Nature, 406, 365.
References ______________________________________________________________ 293
Dixon, M.J ., Smilek, D., Wagar, B.M., Cudahy, C. & Merikle, P.M. (2002). Five plus two
equals yellow: Perceptual and conceptual aspects of synaesthesia. Brain and
Cognition, 48(2-3), 236.
Duffy, P.L. (2001). Blue cats and chartreuse kittens: How synaesthetes color their
world. New York: Henry Holt and Company.
Duncan, J . & Humphreys, G.W. (1989). Visual search and stimulus similarity.
Psychological Review, 96(3), 433-458.
Eagleman, D., Kagan, A., Sarma, K. & Nelson, S. (2006). The genetics of synesthesia:
Linking genes to perception. Oral presentation at the 2
nd
international meeting of
the UK Synaesthesia Association celebrated in London, UK. April 21-23.
Edquist, J ., Rich, A.N., Brinkman, C. & Mattingley, J .B. (2006). Do synaesthetic colours
act as unique features in visual search?. Cortex, 42(2). 222-231.
Elias, L.J , Saucier, D.M, Hardie, C. & Sarty, G.E. (2003). Dissociating semantic and
perceptual components of synaesthesia: Behavioural and functional
neuroanatomical investigations. Cognitive Brain Research, 16, 232-237.
Engel, S.A., Rumelhart, D.E., Wandell, B.A. & Lee, A.T. (1994). fMRI of human visual
cortex. Nature, 369, 525.
Engel, S.A., Zhang, X. & Wandell, B. (1997). Colour tuning in human visual cortex
measured with functional magnetic resonance imaging. Nature, 388, 68-71.
Enns, J . & Di Lollo, V. (1997). Object substitution: a new form of masking in unattended
visual locations. Psychological Science 8, 135139.
Fazio, R.H., Sanbonmatsu, D.M., Powell, M.C. & Kardes, F.R. (1986). On the automatic
activation of attitudes. Journal of Personality and Social Psychology, 50, 229
238.
Fox, E. (1994). Attentional bias in anxiety - a defective inhibition hypothesis. Cognition
& Emotion, 8, 165-195.
Galton, F. (1883). Enquiries into human faculty and its development. London: Dent.
294 _____________________________________________________ REFERENCES
Gevers, W., Reynvoet, B. & Fias, W. (2003). The mental representation of ordinal
sequeences is spatially organized. Cognition, 87, B87-B95.
Goldman-Rakic, P.S., Selemon, L.D. & Schwartz, M.L. (1984). Dual pathways
connecting the dorsolateral prefrontal cortex with the hippocampal-formation and
parahippocampal cortex in the rhesus-monkey, Neuroscience, 12(3), 719-743.
Grossenbacher, P.G. & Lovelace, C.T. (2001). Mechanisms of synesthesia: cognitive
and physiological constraints. Trends in Cognitive Sciences, 5(1), 36-41.
Grossenbacher, P.G. (1997). Perception and sensory information in synaesthetic
experience. In S.Baron-Cohen & J .E.Harrison (Eds.), Synaesthesia: Classic and
contemporary readings. Oxford, England: Blackwell, pp. 148-172.
Harrison, J .E. & Baron-Cohen, S. (1997). Synaesthesia: a review of psychological
theories. In S.Baron-Cohen & J .E.Harrison (Eds.), Synaesthesia: Classic and
Harrison, J .E. (2001). Synaesthesia: the strangest thing. Oxford, England: Oxford
University Press.
Hemphil, M. (1996). A note on adults color-emotion associations. The Journal of
Genetic Psychology, 157(3), 275,280.
Henik, A. & Tzelgov, J . (1982). Is three greater than five: the relation between physical
and semantic size in numerical comparison. Memory and Cognition, 10, 389-395.
Henning, H. (1923). Eine neuartige komplexsynsthesie und komplexzuord-nung.
Zeitschrift fr Psychologie, 92, 149-160.
Hubbard, E.M. & Ramachandran, V.S. (2003). Refining the experimental lever - A reply
to Shanon and Pribram. Journal of Consciousness Studies, 10(3), 77-84.
Hubbard, E.M. & Ramachandran, V.S. (2005). Neurocognitive mechanisms of
synesthesia. Neuron, 48(3), 509-520.
Hubbard, E.M., Arman, A.C, Ramachandran, V.S. & Boynton, G.M. (2005). Individual
differences among grapheme-color synesthetes: brain-behavior correlations.
Neuron, 45, 975-985.
References ______________________________________________________________ 295
Hug-Hellmuth, H. von (1912). ber farbenhren. Imago, 1, 228-264.
Karwoski, T.F., Odbert, H.S. & Osgood, C.E. (1942). Studies in synestehtic thinking II.
The role of form in visual responses to music. Journal of General Psychology, 26,
199-222.
Kennedy, H., Batardiere, A., Dehay, C. & Barone, P. (1997). Synaesthesia: Implications
for developmental neurobiology. In S.Baron-Cohen & J .E.Harrison (Eds.),
Synaesthesia: Classic and contemporary readings. Oxford, England: Blackwell,
pp. 243-256.
Kerns, J .G., Cohen, J .D., Macdonald, A.W., Cho, R.Y., Stenger, V.A. & Carter, C.S.
(2004). Anterior cingulate conflict monitoring and adjustments in control. Science,
303, 1023-1026.
Kim, C-Y., Blake, R. & Palmeri, T.J . (2006). Perceptual interaction between real and
synesthetic colors. Cortex, 42(2), 195-203.
Knoch, D., Gianotti, L.R., Mohr, C. & Brugger, P. (2005). Synesthesia: When colors
count. Brain Research. Cognitive Brain Research, 25(1), 372-374.
Kreitler, H. & Kreitler, S. (1972). Psychology of arts. Durham, NC: Duke University
Press.
Laeng, B., Svartdal, F. & Oelmann, H. (2004). Does color synesthesia pose a paradox
for early-selection theories of attention?. Psychological Science, 15(4), 277-281.
Lakoff, G. & J ohnson, M. (1980). Metaphors we live by. Chicago, IL: The University of
Chicago Press.
Langfield, H.S. (1926). Synesthesia. Psychological Bulletin, 23, 599-602.
Lazarus, R.S. (1991). Emotion and adaptation. New York: Oxford University Press.
LeDoux, J .E. (1992). Brain mechanisms of emotion and emotional learning. Current
Opinion in Neurobiology, 2(2), 191-197.
Li, Y. (1996). A study on negative priming effect in selective attention. Acta
Psychologica Sinica, 28(3), 245-251.
296 _____________________________________________________ REFERENCES
Lupiez, J . & Callejas, A. (2006). Automatic perception and synesthesia: evidence
from color and photism naming in a Stroop-negative priming task. Cortex, 42(2),
204-212.
Luria, A.R. (1968). The mind of a mnemonist. Cambridge, MA: Harvard University
Press.
MacLeod, C.M. (1991). Half a century of research on the Stroop effect: an integrative
review. Psychologial Bulletin, 109, 163203.
Maddock, R.J . (1999) Retrosplenial cortex and emotion: New insights from functional
imaging studies of the human brain. Trends in Neurosciences, 22, 310316.
Maddock, R.J ., Garrett, A.S. & Buonocore, M.H. (2003). Posterior cingulate cortex
activation by emotional words: fmri evidence from a valence decision task.
Human Brain Mapping, 18, 3041.
Mandler, G. (1984). Mind and Body. Psychology of Emotion and Stress. New York:
Norton and Company.
Mari-Beffa, P., Fuentes, L.J ., Catena, A., Houghton, G. (2000). Semantic priming in the
prime task effect: evidence of automatic semantic processing of distractors.
Memory and Cognition, 28(4),635-647.
Marks, L.E. & Odgaard, E.C. (2005). Developmental constraints on theories of
synesthesia. In L.C.Robertson & N.Savig (Eds.), Synesthesia: Perspectives from
Cognitive Neuroscience. New York: Oxford University Press, pp. 214-236.
Marks, L.E. (1975). On colored-hearing synesthesia: cross-modal translations of
sensory dimensions. Psychological Bulletin, 82(3), 303-331.
Mattingley, J .B. & Rich, A.N. (2004). Behavioral and brain correlates of multisensory
experience in synaesthesia. In G.A.Calvert, C.Spence, & B.E.Stein (Eds.),
Handbook of multisensory processes. Cambridge, MA: MIT Press, pp. 851-865.
Mattingley, J .B., Payne, J .M. & Rich, A.N. (2006). Attentional load attenuates
synaesthetic priming effects in grapheme-colour synaesthesia. Cortex, 42(2),
213-221.
References ______________________________________________________________ 297
Mattingley, J .B., Rich, A.N., Yelland, G. & Bradshaw, J .L. (2001). Uncosncious priming
eliminates automatic binding of colour and alphanumeric form in synaesthesia.
Nature 410, 580-582.
Maurer, D. & Mondloch, C.J . (1996). Synesthesia: a stage of normal infancy?. In
S.Masin (Ed.), Proceedings of the 12
th
meeting of the international society of
psychophysics (pp. 107-112). Padua.
Maurer, D. & Mondloch, C.J . (2005). Neonatal synesthesia: areevaluation. In
Maurer, D. (1997). Neonatal synaesthesia: implications for the processing of speech
and faces. In S.Baron-Cohen & J .E.Harrison (Eds.), Synaesthesia: Classic and
Meier & Robinson (2004) Why the sunny side is up: Associations between affect and
vertical position. Psychological Science, 15, 243-247.
Mills, C.B., Boteler, E.H. & Oliver, G.K. (1999). Digit synaesthesia: A case study using a
Stroop-type test. Cognitive Neuropsychology, 16(2), 181-191.
Mills, C.B., J oanne, I., Westendorf, T, Owsianiecki, L. & McDonald, A. (2006). Effect of
a synesthetes photisms on name recall. Cortex, 42(2), 155-163.
Mills, C.B., Viguers, M.L., Edelson, S.K., Thomas, A.T., Simon-Dack, S.L. & Innis, J .A.
(2002). The color of two alphabets for a multilingual synesthete. Perception, 31,
1371-1394.
Mitchell, K.J ., Bargary, G., Barnett, K.J . & Newell, F.N. (2006). Synaesthesia and
principles of cortical connectivity. Oral presentation at the 2
nd
international
meeting of the UK Synaesthesia Association celebrated in London, UK. April 21-
23.
Mulvenna, C., Hubbard, E.M., Ramachandran, V.S., and Pollick, F. (2004). The
relationship between synaesthesia and creativity. Journal of Cognitive
Neuroscience. Suppl. 16, 188.
298 _____________________________________________________ REFERENCES
Myles, K.M, Dixon, M.J ., Smilek, D. & Merikle, P.M. (2003). Seeing double: the role of
meaning in alphanumeric-color synaesthesia. Brain and Cognition, 53, 342-345.
Navon, D. (1977). Forest before trees: The precedence of global features in visual
perception. Cognitive Psychology, 9, 353-383.
Neill, W.T. & Westberry, W.L.(1987). Selective attention and the suppression of
cognitive noise. Journal of Experimental Psychology-Learning Memory and
Cognition, 13, 327-334.
Newton, I. (1730). Optiks (4 ed., 1952). New York: Dover.
Nunn, J .A., Gregory, L.J ., Brammer, M., Williams, S.C.R., Parslow, D.M., Morgan, M.J .,
Morris, R.G., Bullmore, E.T., Baron-Cohen, S. & Gray, J .A. (2002). Functional
magnetic resonance imaging of synesthesia: Activation of V4/V8 by spoken
words. Nature Neuroscience, 5, 371375.
Odbert, H.S., Karwoski, T.F. & Eckerson, A.B. (1942). Studies in synesthetic thinking I.
Musical and verbal association of color and mood. Journal of General
Psychology, 26, 153-173.
Odgaard, E.C., Flowers, J .H. & Bradman, H.L. (1999). An investigation of the cognitive
and perceptual dynamics of a colour-digit synaesthete. Perception, 28(5), 651-
664.
Olson, M.A. & Fazio, R.H. (2001). Implicit attitude formation through classical
conditioning. Psychological Science, 12(5), 413-417.
Ortigue, S., Michael, C.M., Murray, M.M., Mohr, C., Carbonnel, S. & Landis, T. (2004).
Electrical neuroimaging reveals early generator modulation to emotional words.
NeuroImage, 21, 1242-1251.
Palmeri, T.J ., Blake, R., Marois, R., Flanery, M.A. & Whetsell J r, W. (2002). The
perceptual reality of synesthetic colors. Proceedings of the National Academy of
Sciences, 99(6), 4127-4131.
Paulesu, E., Harrison, J ., Baron-Cohen, S., Watson, J . D., Goldstein, L., Heather, J ., et
al. (1995). The physiology of coloured hearing. A PET activation study of colour-
word synaesthesia. Brain, 118(3), 661-676.
References ______________________________________________________________ 299
Pulvermller, F. (2001). Brain reflections of words and their meaning. Trends in
Cognitive Sciences, 5, 517524.
Ramachandran, V.S. & Hubbard, E.M. (2001a). Psychophysical investigations into the
neural basis of synaesthesia. Proceedings of the Royal Society of London Series
B Biological Sciences, 268, 979-983.
Ramachandran, V.S. & Hubbard, E.M. (2001b). Synesthesia- A window into perception,
thought and language. Journal of Consciousness Studies. 8(12), 3-34
Razran, G.H.S. (1938). Conditioning away social bias by the luncheon thechnique.
Psychological Bulletin, 37, 481.
Reichard, G., J akobson, R & Werth, E. (1949). Language and synesthesia. Word, 5,
224-233.
Rich, A.N. & Mattingley, J .B. (2002). Anomalous perception in synaesthesia: A cognitive
neuroscience perspective. Nature Reviews Neuroscience 3(1), 43-52.
Rich, A.N. & Mattingley, J .B. (2003). The effects of stimulus competition and voluntary
attention on colour-graphemic synaesthesia. Neuroreport 14(14), 1793-1797.
Rich, A.N. & Mattingley, J .B. (2005). Can attention modulate color-graphemic
synesthesia?. In L.C.Robertson & N.Savig (Eds.), Synesthesia: Perspectives from
Cognitive Neuroscience. New York: Oxford University Press, pp. 108-123.
Rich, A.N., Bradshaw, J .L. & Mattingley, J .B. (2005). A systematic, large-scale study of
synaesthesia: implications for the role of early experience in lexical-colour
associations. Cognition, 98(1), 53-84.
Robertson L.C. & Savig N. (2005). Synesthesia: Perspectives from Cognitive
Neuroscience. New York: Oxford University Press .
Robertson, L.C. (2001). Color my is blue. Nature, 410, 533-534.
Robertson, L.C. (2003). Binding, spatial attention and perceptual awareness. Nature
Reviews Neuroscience, 4, 93102.
Sagiv N. & Ward, J . (in press). Cross-modal interactions: Lessons from synesthesia. In
S.Martinez-Conde, L.M.Martinez, J .M.Alonso, P.U.Tse, P.U. & S.L.Macknik (Eds),
300 _____________________________________________________ REFERENCES
Progress in Brain Research Series, Vol. 155: Visual Perception. London: Elsevier
Science.
Sagiv, N. & Robertson, L.C. (2005). Synesthesia and the binding problem. In
Sagiv, N., Heer, J . & Robertson, L.C. (2006). Does binding of synesthetic color to the
evoking grapheme require attention?. Cortex, 42(2), 232-242.
Schiltz, K., Trocha, K., Wieringa, B.M., Emrich, H.M., J ohannes, S. & Munte, T.F.
(1999). Neurophysiological aspects of synesthetic experience. Journal of
Neuropsychiatry and Clinical Neurosciences, 11(1), 58-65.
Schneider, W. (1988). Micro Experimental Laboratory: An integrated system for IBM PC
compatibles. Behavior Research Methods, Instruments and Computers, 20(2),
206-217.
Schneider, W., Eschman, A. & Zuccolotto, A. (2002) E-Prime users guide. Psychology
Software Tools Inc., Pittsburgh, PA
Sereno, M.I., Dale, A.M., Reppas, J .B. & Kwong, K.K. (1995). Borders of multiple visual
areas in humans revealed by functional magnetic resonance imaging. Science,
268, 889-893.
Seron, X., Psenti, M. Nol, M.P., Deloche, G. y Cornet, J .A. (1992). Images of numbers,
or when 98 is upper left and 6 sky blue. Cognition, 44, 159-196.
Simner, J ., Glover, L. & Mowat, A. (2006). Linguistic determinats of word colouring in
grapheme-colour synaesthesia. Cortex, 42(2), 281-289.
Simner, J ., Lanz, M., J ansari, A., Noonan, K., Glover, L., Oakley, D.A. & Ward, J .
(2005). Non-random associations of graphemes to colos in synaesthetic and
normal populations. Cognitive Neuropsychology, 22(8), 1069-1085.
Simner, J ., Mulvenna, C., Sagiv, N. Tsakanikos, E., Witherby, S.A., Fraser, C. Scott, K
& Ward J . (in press). Synaesthesia: The prevalence of atypical cross-modal
experiences. Perception
References ______________________________________________________________ 301
Smilek, D. & Dixon, M.J . (2002). Towards a synergistic understanding of synaesthesia:
combining current experimental findings with synaesthetes' subjective
descriptions. Psyche 8(1) (http://psyche.cs.monash.edu.au/v8/psyche-8-01-
smilek.html)
Smilek, D. Dixon, M.J . & Merikle, P.M. (2005a). Binding of graphemes and synesthetic
colors in color-graphemic synesthesia. In L.C.Robertson & N.Savig (Eds.),
Synesthesia: Perspectives from Cognitive Neuroscience. New York: Oxford
University Press, pp. 74-89.
Smilek, D., Callejas, A., Dixon, M.J . & Merikle, P.M. (in press). Ovals of time: time-
space relationships in synaesthesia. Consciousness and Cognition.
Smilek, D., Dixon, M.J . & Merikle, P.M. (2005b). Synaesthesia: discordant male
monozygotic twins. Neurocase, 11(5), 363-370.
Smilek, D., Dixon, M.J ., Cudahy, C. & Merikle, P.M. (2001). Synaesthetic photisms
influence visual perception. Journal of Cognitive Neuroscience 13(7), 930-936.
Smilek, D., Dixon, M.J ., Cudahy, C. & Merikle, P.M. (2002). Synesthetic color
experiences influence memory. Psychological Science, 13(6), 548-552.
Smilek, D., Dixon, M.J ., Cudahy, C., & Merikle, P.M. (2002).Concept driven color
experiences in digit-color synesthesia. Brain and Cognition, 48, 570-573.
Smilek, D., Moffatt, B.A., Pasternak, J ., White, B.N., Dixon, M.J . & Merikle, P.M. (2002).
Synaesthesia: A case study of discordant monozygotic twins. Neurocase, 8(4),
338-342.
Smith, H.L. (1905). Synesthesia. Bulletin of the Johns Hopkins Hospital, 16, 258-263.
Staats, A.W. & Staats, C.K. (1958). Attitudes established by classical conditioning.
Journal of Abnormal and Social Psychology, 11, 187-192.
Stroop, J .R. (1935). Studies of interferente in serial verbal reactions. Journal of
Experimental Psychology, 18, 643-662.
Suarez de Mendoza, F. (1890). Laudition colore. Paris: Octave Donin.
302 _____________________________________________________ REFERENCES
Terwogt, M.M. & Hoeksma, J .B. (1995). Colors and emotions: preferences and
combinations. The Journal of General Psychology, 122(1), 5-17.
Tipper, S.P. (1985). The negative priming effect - inhibitory priming by ignored objects.
Quarterly Journal of Experimental Psychology Section A-Human Experimental
Psychology, 37, 571-590.
Tipper, S.P. (2001). Does negative priming reflect inhibitory mechanisms? A review and
integration of conflicting views. Quarterly Journal of Experimental Psychology
Section A-Human Experimental Psychology, 54, 321-343.
Treisman, A.M. & Gelade, G. (1980). A feature-integration theory of attention. Cognitive
psychology, 12, 97-136.
Van Hoesen, G.W., Morecraft, R.J . & Vogt, B.A. (1993). Connections of the monkey
cingulate cortex. In Vogt & Gabriel, (Eds.), Neurobiology of cingulate cortex and
limbic thalamus. Birkhuser, pp. 249-284
Wade, A.R., Brewer, A.A., Rieger, J .W., & Wandell, B.A. (2002). Functional
measurements of human ventral occipital cortex: retinotopy and colour.
Philosophical Transactions of the Royal Society of London B. Biological Sciences,
357, 963973.
Wagar, B.M., Dixon, M.J ., Smilek, D. & Cudahy, C. (2002). Colored photisms prevent
object-substitution masking in digit-colour synesthesia. Brain and Cognition, 48,
606-611.
Ward, J . & Mattingley, J .B. (2006). Synaesthesia: an overview of contemporary findings
and controversies. Cortex, 42(2), 129-136.
Ward, J . & Simner, J . (2003). Lexical-gustatory synaesthesia: linguistic and conceptual
factors. Cognition, 89(3), 237-261.
Ward, J . & Simner, J . (2005). Is synaesthesia an X-linked dominant trait with lethality in
males? Perception, 34(5), 611-623.
Ward, J . (2004). Emotionally mediated synaesthesia. Cognitive neuropsychology,
21(7), 761772.
References ______________________________________________________________ 303
Ward, J ., Huckstep, B. & Tsakanikos, E. (2006). Sound-color synaesthesia: to what
extend does it use cross.modal mechanisms common to us all?. Cortex, 42(2),
264-280.
Ward, J ., Li, R., Salih, S. & Sagiv, N. (2006). Whats right and whats wrong with the
projector-associator distinction?. Oral presentation at the 2
nd
UK Synesthesia
Association Meeting, London, April 21
st
-23
rd
.
Weiss, P.H., Shah, N.J ., Toni, I., Zilles, K. & Fink, G.R. (2001). Associating colours with
people: A case of chromatic-lexical synaesthesia. Cortex, 37(5), 750-753.
Weiss, P.H., Zilles, K., and Fink, G.R. (2005). When visual perception causes feeling:
enhanced cross-modal processing in graphemecolor synesthesia. Neuroimage,
28, 859-868.
Wheeler, R.H. & Cutsforth, T.D. (1922). Synesthesia, a form of perception. Psychology
Review, 29, 212-220.
Witthoft, N. & Winawer, J . (2006). Synesthetic colors determined by having colored
refrigetaror magnets in childhood. Cortex, 42(2), 175-183.
Wollen, K.A. & Ruggiero, F.T. (1983). Coloured-letter synaesthesia. Journal of Mental
Imagery, 7, 83-86.
Zeki, S. & Marini, L. (1998). Three cortical stages of colour processing in the human
brain. Brain, 121(9), 1669-1685.

16134540

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

16134540

Uploaded by

Copyright:

Available Formats

TESIS DOCTORAL

You might also like